fix(models): route auto-discovered providers by wire protocol (#41)

- replace npmToLLMProvider map with npmToWireProtocol (openai/anthropic/google)
- add createAutoRoutedGoogleProvider so @ai-sdk/google proxies work
  (fixes opencode/gemini-* failing with "no LLM provider mapping")
- strip the genai-injected v1beta segment for proxies whose base URL
  already carries a version (e.g. opencode's /zen/v1)
- preserve openai-compat fallback and clearer error for unroutable providers
- document auto-routing in README and providers docs; update CreateProvider godoc
- add regression tests for wire routing and version-path rewriting

Fixes #41

.kit/extensions/go-edit-lint.go

@@ -1,268 +0,0 @@
-//go:build ignore
-
-package main
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"os/exec"
-	"path/filepath"
-	"strings"
-	"time"
-
-	"kit/ext"
-)
-
-const (
-	diagnosticsTimeout = 20 * time.Second
-	maxOutputBytes     = 12_000
-)
-
-type toolPathInput struct {
-	Path string `json:"path"`
-}
-
-type lintResult struct {
-	Output string
-	Err    error
-}
-
-// Package-level state: set of .go files edited during the current agent turn.
-var editedFiles map[string]bool
-
-func Init(api ext.API) {
-	api.OnSessionStart(func(_ ext.SessionStartEvent, ctx ext.Context) {
-		ctx.Print("go-edit-lint extension loaded - will run gopls and golangci-lint after agent turns that edit Go files")
-	})
-
-	// Track edited .go files — don't lint yet.
-	api.OnToolResult(func(e ext.ToolResultEvent, ctx ext.Context) *ext.ToolResultResult {
-		if e.IsError || !isEditOrWrite(e.ToolName) {
-			return nil
-		}
-
-		absPath, ok := resolveGoFilePath(e.Input, ctx.CWD)
-		if !ok {
-			return nil
-		}
-
-		if editedFiles == nil {
-			editedFiles = make(map[string]bool)
-		}
-		editedFiles[absPath] = true
-		return nil
-	})
-
-	// After the agent turn ends, lint all collected files.
-	api.OnAgentEnd(func(e ext.AgentEndEvent, ctx ext.Context) {
-		if len(editedFiles) == 0 {
-			return
-		}
-
-		// Snapshot and reset immediately so the next turn starts clean.
-		files := editedFiles
-		editedFiles = nil
-
-		// Skip lint on errored turns.
-		if e.StopReason == "error" {
-			return
-		}
-
-		// Collect unique directories and file list for gopls.
-		var allGoplsOutput []string
-		for absPath := range files {
-			res := runGopls(ctx.CWD, absPath)
-			formatted := formatToolResult(res, "")
-			if formatted != "" {
-				allGoplsOutput = append(allGoplsOutput, fmt.Sprintf("# %s\n%s", filepath.Base(absPath), formatted))
-			}
-		}
-
-		lintRes := runGolangCILint(ctx.CWD, "./...")
-
-		goplsSection := "No diagnostics."
-		if len(allGoplsOutput) > 0 {
-			goplsSection = strings.Join(allGoplsOutput, "\n\n")
-		}
-		lintSection := formatToolResult(lintRes, "No lint issues.")
-
-		// Build file list for the report header.
-		var fileNames []string
-		for absPath := range files {
-			fileNames = append(fileNames, filepath.Base(absPath))
-		}
-
-		report := fmt.Sprintf(
-			"<go_diagnostics files=%q>\n[gopls]\n%s\n\n[golangci-lint]\n%s\n</go_diagnostics>",
-			strings.Join(fileNames, ", "),
-			goplsSection,
-			lintSection,
-		)
-
-		goplsIssues, lintIssues := countIssues(report)
-		hasIssues := goplsIssues > 0 || lintIssues > 0
-
-		if hasIssues {
-			// Show TUI block so the user sees it too.
-			var msgLines []string
-			msgLines = append(msgLines, fmt.Sprintf("Files: %s", strings.Join(fileNames, ", ")))
-			if goplsIssues > 0 {
-				msgLines = append(msgLines, fmt.Sprintf("gopls: %d issue(s)", goplsIssues))
-			}
-			if lintIssues > 0 {
-				msgLines = append(msgLines, fmt.Sprintf("golangci-lint: %d issue(s)", lintIssues))
-			}
-
-			borderColor := "#f9e2af" // yellow
-			if goplsIssues > 0 && lintIssues > 0 {
-				borderColor = "#f38ba8" // red
-			}
-
-			ctx.PrintBlock(ext.PrintBlockOpts{
-				Text:        strings.Join(msgLines, "\n"),
-				BorderColor: borderColor,
-				Subtitle:    "go-edit-lint",
-			})
-
-			// Inject a follow-up message so the agent fixes the issues.
-			ctx.SendMessage(report + "\n\n⚠️ DIAGNOSTICS FOUND: Please review and fix the issues above.")
-		} else {
-			ctx.PrintBlock(ext.PrintBlockOpts{
-				Text:        fmt.Sprintf("Files: %s\n✓ All clean", strings.Join(fileNames, ", ")),
-				BorderColor: "#a6e3a1",
-				Subtitle:    "go-edit-lint",
-			})
-		}
-	})
-}
-
-func isEditOrWrite(toolName string) bool {
-	return strings.EqualFold(toolName, "edit") || strings.EqualFold(toolName, "write")
-}
-
-func resolveGoFilePath(inputJSON, cwd string) (string, bool) {
-	var args toolPathInput
-	if err := json.Unmarshal([]byte(inputJSON), &args); err != nil || args.Path == "" {
-		return "", false
-	}
-
-	absPath := args.Path
-	if !filepath.IsAbs(absPath) {
-		absPath = filepath.Join(cwd, absPath)
-	}
-
-	if strings.ToLower(filepath.Ext(absPath)) != ".go" {
-		return "", false
-	}
-
-	return absPath, true
-}
-
-func runGopls(cwd, absPath string) lintResult {
-	ctx, cancel := context.WithTimeout(context.Background(), diagnosticsTimeout)
-	defer cancel()
-
-	cmd := exec.CommandContext(ctx, "gopls", "check", absPath)
-	cmd.Dir = cwd
-	out, err := cmd.CombinedOutput()
-
-	if ctx.Err() == context.DeadlineExceeded {
-		return lintResult{Err: fmt.Errorf("timed out after %s", diagnosticsTimeout)}
-	}
-
-	if err != nil {
-		return lintResult{Output: truncate(string(out), maxOutputBytes), Err: fmt.Errorf("failed to run gopls check: %w", err)}
-	}
-
-	return lintResult{Output: truncate(string(out), maxOutputBytes)}
-}
-
-func runGolangCILint(cwd, target string) lintResult {
-	ctx, cancel := context.WithTimeout(context.Background(), diagnosticsTimeout)
-	defer cancel()
-
-	args := []string{
-		"run",
-		target,
-		"--show-stats=false",
-		"--output.text.path", "stdout",
-		"--output.text.colors=false",
-		"--output.text.print-issued-lines=false",
-	}
-	cmd := exec.CommandContext(ctx, "golangci-lint", args...)
-	cmd.Dir = cwd
-	out, err := cmd.CombinedOutput()
-
-	if ctx.Err() == context.DeadlineExceeded {
-		return lintResult{Err: fmt.Errorf("timed out after %s", diagnosticsTimeout)}
-	}
-
-	trimmed := truncate(string(out), maxOutputBytes)
-	if err == nil {
-		return lintResult{Output: trimmed}
-	}
-
-	exitErr, ok := err.(*exec.ExitError)
-	if ok && exitErr.ExitCode() == 1 {
-		return lintResult{Output: trimmed}
-	}
-
-	return lintResult{Output: trimmed, Err: fmt.Errorf("failed to run golangci-lint: %w", err)}
-}
-
-func formatToolResult(res lintResult, emptyFallback string) string {
-	var lines []string
-	if res.Err != nil {
-		lines = append(lines, "ERROR: "+res.Err.Error())
-	}
-	out := strings.TrimSpace(res.Output)
-	if out == "" {
-		if res.Err == nil {
-			if emptyFallback != "" {
-				lines = append(lines, emptyFallback)
-			}
-		}
-	} else {
-		lines = append(lines, out)
-	}
-	if len(lines) == 0 {
-		return emptyFallback
-	}
-	return strings.Join(lines, "\n")
-}
-
-func truncate(s string, max int) string {
-	if len(s) <= max {
-		return s
-	}
-	return s[:max] + "\n... output truncated ..."
-}
-
-func countIssues(report string) (goplsCount, lintCount int) {
-	goplsStart := strings.Index(report, "[gopls]")
-	lintStart := strings.Index(report, "[golangci-lint]")
-	endTag := strings.Index(report, "</go_diagnostics>")
-
-	if goplsStart != -1 && lintStart != -1 {
-		goplsSection := report[goplsStart:lintStart]
-		for _, line := range strings.Split(goplsSection, "\n") {
-			line = strings.TrimSpace(line)
-			if line != "" && line != "[gopls]" && line != "No diagnostics." && !strings.HasPrefix(line, "#") {
-				goplsCount++
-			}
-		}
-	}
-
-	if lintStart != -1 && endTag != -1 {
-		lintSection := report[lintStart:endTag]
-		for _, line := range strings.Split(lintSection, "\n") {
-			line = strings.TrimSpace(line)
-			if line != "" && line != "[golangci-lint]" && line != "No lint issues." {
-				lintCount++
-			}
-		}
-	}
-
-	return goplsCount, lintCount
-}

README.md

@@ -127,6 +127,7 @@ max-tokens: 4096
 temperature: 0.7
 stream: true
 thinking-level: off       # off, none, minimal, low, medium, high
+no-core-tools: false      # set to true to disable all built-in core tools
 ```
 
 All of the above keys can also be set programmatically via the SDK
@@ -195,9 +196,10 @@ mcpServers:
 --compact                Enable compact output mode
 --auto-compact           Auto-compact conversation near context limit
 
-# Extensions
+# Extensions and tools
 --extension, -e          Load additional extension file(s) (repeatable)
 --no-extensions          Disable all extensions
+--no-core-tools          Disable all built-in core tools (bash, read, write, edit, grep, find, ls, subagent)
 --prompt-template        Load a specific prompt template by name
 --no-prompt-templates    Disable prompt template loading
 
@@ -554,7 +556,7 @@ host, err := kit.New(ctx, &kit.Options{
     SystemPrompt: "You are a helpful bot",
     ConfigFile:   "/path/to/config.yml",
     MaxSteps:     10,
-    Streaming:    true,
+    Streaming:    ptr(true), // *bool: nil = unset (default true), &false = off
     Quiet:        true,
 
     // Generation parameters (override env/config/per-model defaults)
@@ -579,7 +581,9 @@ host, err := kit.New(ctx, &kit.Options{
     // Tool options
     Tools:            []kit.Tool{...},     // Replace default tool set entirely
     ExtraTools:       []kit.Tool{...},     // Add tools alongside defaults
-    DisableCoreTools: true,                // Use no core tools (0 tools, for chat-only)
+    DisableCoreTools: true,                // Disable all built-in core tools; also controllable via
+                                           // --no-core-tools flag, KIT_NO_CORE_TOOLS env var,
+                                           // or no-core-tools: true in .kit.yml
 
     // Configuration
     SkipConfig:   true,                   // Skip .kit.yml files (viper defaults + env vars still apply)
@@ -599,6 +603,38 @@ are pointer types so explicit `0.0` is distinguishable from "leave alone"; a
 non-zero `MaxTokens` suppresses automatic right-sizing the same way `--max-tokens`
 does on the CLI.
 
+### Functional options (`NewAgent`)
+
+For simple programmatic setups, `kit.NewAgent` offers an ergonomic
+functional-options front door over `kit.New`. Streaming is **enabled by
+default**; pass `kit.WithStreaming(false)` to opt out.
+
+```go
+host, err := kit.NewAgent(ctx,
+    kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+    kit.WithSystemPrompt("You are a helpful assistant."),
+    kit.WithMaxTokens(8192),
+    kit.WithThinkingLevel("medium"),
+    kit.Ephemeral(), // in-memory session, no persistence
+)
+```
+
+Available options: `WithModel`, `WithSystemPrompt`, `WithStreaming`,
+`WithMaxTokens`, `WithThinkingLevel`, `WithTools`, `WithExtraTools`,
+`WithProviderAPIKey`, `WithProviderURL`, `WithConfigFile`, `WithDebug`, and
+`Ephemeral`. For advanced configuration not covered by the helpers (custom MCP
+config, in-process MCP servers, session backends, MCP task tuning) construct an
+`Options` value explicitly and call `kit.New`.
+
+### Per-instance config isolation
+
+Each `kit.New` / `kit.NewAgent` call owns an **isolated configuration store**,
+so constructing multiple Kit instances in the same process is safe: setting the
+model, thinking level, or generation parameters on one never affects another,
+and runtime mutators (`SetModel`, `SetThinkingLevel`) only touch the owning
+instance. This makes subagent spawning and multi-Kit embedding race-free with
+no external synchronization required.
+
 ### MCP OAuth (remote MCP servers)
 
 When a remote MCP server returns 401, Kit runs the full OAuth flow (dynamic
@@ -756,6 +792,45 @@ host, _ := kit.New(ctx, &kit.Options{
 })
 ```
 
+### Runtime Skills & Context Files
+
+For multi-tenant hosts (chatbots, per-user agents, web services), the SDK
+lets you swap skills and `AGENTS.md`-style context files **after** Kit
+construction. Every mutation recomposes the system prompt and applies it to
+the agent so the next turn picks up the new instructions — no restart needed.
+
+```go
+// Programmatic skill (no file on disk required).
+host.AddSkill(&kit.Skill{
+    Name:        "polite-french",
+    Description: "Respond in French and always greet the user.",
+    Content:     "Always reply in French. Open every response with 'Bonjour'.",
+})
+
+// Or load one from disk.
+host.LoadAndAddSkill("/var/skills/refund-policy.md")
+
+// Per-user AGENTS.md content pulled from a database.
+host.AddContextFileContent(
+    fmt.Sprintf("session://%s/AGENTS.md", userID),
+    rulesFromDB,
+)
+
+// Tear down session-specific state on logout.
+host.RemoveSkill("polite-french")
+host.RemoveContextFile(fmt.Sprintf("session://%s/AGENTS.md", userID))
+
+// Or replace the whole set atomically.
+host.SetSkills(activeSkillsForUser)
+host.SetContextFiles(activeContextForUser)
+```
+
+Skills dedupe by `Name`, context files dedupe by `Path` (which can be any
+opaque identifier — it doesn't have to be a real filesystem path). All
+mutators and readers (`GetSkills`, `GetContextFiles`) are safe to call
+concurrently from multiple goroutines. See the [SDK overview docs](/sdk/overview#runtime-skills-and-context-files)
+for the full reference.
+
 ## Advanced Usage
 
 ### Subagent Pattern
@@ -897,6 +972,31 @@ This automatically defaults to `custom/custom` without needing to specify a mode
 - Reasoning and temperature support
 - Optional `CUSTOM_API_KEY` environment variable or `--provider-api-key` flag
 
+### Auto-routed Providers
+
+Any provider in the [models.dev](https://models.dev) database can be used as
+`provider/model` without a dedicated native integration. Kit auto-routes the
+request through the matching **wire protocol** based on the provider's npm package
+(or per-model override), using its `api` URL as the base:
+
+| npm package | Wire protocol |
+|-------------|---------------|
+| `@ai-sdk/openai` | OpenAI (Responses API) |
+| `@ai-sdk/openai-compatible` | OpenAI (chat completions) |
+| `@ai-sdk/anthropic` | Anthropic |
+| `@ai-sdk/google` | Google Gemini |
+
+Providers with an `api` URL but an unrecognized npm package fall back to the
+OpenAI-compatible wire. Because routing follows the wire protocol, aggregator/proxy
+providers work across all of their models — including Claude, GPT, *and* Gemini
+routes:
+
+```bash
+kit --model opencode/claude-haiku-4-5 "Hello"     # → Anthropic wire
+kit --model opencode/gpt-5 "Hello"                # → OpenAI wire
+kit --model opencode/gemini-3.5-flash "Hello"     # → Google wire
+```
+
 ### Model String Format
 
 ```bash

cmd/root.go

@@ -71,6 +71,7 @@ var (
 
 	// Extensions control
 	noExtensionsFlag bool
+	noCoreToolsFlag  bool
 	extensionPaths   []string
 
 	// TLS configuration
@@ -278,6 +279,8 @@ func init() {
 		BoolVar(&noSessionFlag, "no-session", false, "ephemeral mode — no session persistence")
 	rootCmd.PersistentFlags().
 		BoolVar(&noExtensionsFlag, "no-extensions", false, "disable all extensions")
+	rootCmd.PersistentFlags().
+		BoolVar(&noCoreToolsFlag, "no-core-tools", false, "disable all built-in core tools (bash, read, write, edit, grep, find, ls, subagent)")
 	rootCmd.PersistentFlags().
 		StringSliceVarP(&extensionPaths, "extension", "e", nil, "load additional extension file(s)")
 
@@ -327,6 +330,7 @@ func init() {
 	_ = viper.BindPFlag("main-gpu", rootCmd.PersistentFlags().Lookup("main-gpu"))
 	_ = viper.BindPFlag("tls-skip-verify", rootCmd.PersistentFlags().Lookup("tls-skip-verify"))
 	_ = viper.BindPFlag("no-extensions", rootCmd.PersistentFlags().Lookup("no-extensions"))
+	_ = viper.BindPFlag("no-core-tools", rootCmd.PersistentFlags().Lookup("no-core-tools"))
 	_ = viper.BindPFlag("extension", rootCmd.PersistentFlags().Lookup("extension"))
 	_ = viper.BindPFlag("prompt-template", rootCmd.PersistentFlags().Lookup("prompt-template"))
 	_ = viper.BindPFlag("no-prompt-templates", rootCmd.PersistentFlags().Lookup("no-prompt-templates"))
@@ -434,6 +438,36 @@ func extensionCommandsForUI(k *kit.Kit) []commands.ExtensionCommand {
 	return cmds
 }
 
+// buildExtensionItems converts the loaded extensions into ui.ExtensionItem
+// values for the startup [Extensions] section. The display name is the file
+// basename without the .go suffix; for subdirectory extensions the parent
+// directory name is used (matching `kit extensions list`).
+func buildExtensionItems(k *kit.Kit, cwd string) []ui.ExtensionItem {
+	infos := k.Extensions().Loaded()
+	if len(infos) == 0 {
+		return nil
+	}
+	items := make([]ui.ExtensionItem, 0, len(infos))
+	for _, info := range infos {
+		name := filepath.Base(info.Path)
+		if name == "main.go" {
+			// Subdirectory extension: use the parent directory's name.
+			name = filepath.Base(filepath.Dir(info.Path))
+		}
+		name = strings.TrimSuffix(name, ".go")
+		source := "user"
+		if cwd != "" && strings.HasPrefix(info.Path, cwd) {
+			source = "project"
+		}
+		items = append(items, ui.ExtensionItem{
+			Name:   name,
+			Path:   info.Path,
+			Source: source,
+		})
+	}
+	return items
+}
+
 // widgetProviderForUI returns a function that converts extension widgets to
 // ui.WidgetData for the given placement. Returns nil if extensions are
 // disabled, which is safe — the UI treats a nil GetWidgets as "no widgets".
@@ -742,13 +776,14 @@ func runNormalMode(ctx context.Context) error {
 	var appInstancePtr *app.App
 
 	kitOpts := &kit.Options{
-		Quiet:          quietFlag,
-		Debug:          debugMode,
-		NoSession:      noSessionFlag,
-		Continue:       continueFlag,
-		SessionPath:    sessionPath,
-		AutoCompact:    autoCompactFlag,
-		MCPAuthHandler: authHandler,
+		Quiet:            quietFlag,
+		Debug:            debugMode,
+		NoSession:        noSessionFlag,
+		Continue:         continueFlag,
+		SessionPath:      sessionPath,
+		AutoCompact:      autoCompactFlag,
+		MCPAuthHandler:   authHandler,
+		DisableCoreTools: viper.GetBool("no-core-tools"),
 		// This callback is called when each MCP server finishes loading.
 		// We use a closure that captures appInstancePtr which is set after
 		// app.New() is called below.
@@ -869,8 +904,9 @@ func runNormalMode(ctx context.Context) error {
 			appInstance:  appInstance,
 			usageTracker: usageTracker,
 		})
+
+		// During startup, buffer extension messages so they appear after the banner.
 		extCtx.Print = func(text string) {
-			// Capture messages during startup, print after startup banner.
 			startupExtensionMessages = append(startupExtensionMessages, text)
 		}
 		extCtx.PrintInfo = func(text string) {
@@ -883,15 +919,6 @@ func runNormalMode(ctx context.Context) error {
 		kitInstance.Extensions().EmitSessionStart()
 
 		// Restore normal print functions for runtime use.
-		extCtx = buildInteractiveExtensionContext(extensionContextDeps{
-			ctx:          ctx,
-			cwd:          cwd,
-			modelName:    modelName,
-			interactive:  positionalPrompt == "",
-			kitInstance:  kitInstance,
-			appInstance:  appInstance,
-			usageTracker: usageTracker,
-		})
 		extCtx.Print = func(text string) { appInstance.PrintFromExtension("", text) }
 		extCtx.PrintInfo = func(text string) { appInstance.PrintFromExtension("info", text) }
 		extCtx.PrintError = func(text string) { appInstance.PrintFromExtension("error", text) }
@@ -942,6 +969,10 @@ func runNormalMode(ctx context.Context) error {
 		})
 	}
 
+	// Build extension items from the loaded extensions for the [Extensions]
+	// startup section. Each entry is a single .go file (or a subdir's main.go).
+	extensionItems := buildExtensionItems(kitInstance, cwd)
+
 	// Build prompt template and skill item provider callbacks for hot-reload.
 	// These are called by the TUI when ContentReloadEvent fires.
 	getPromptTemplates := func() []*prompts.PromptTemplate {
@@ -986,6 +1017,13 @@ func runNormalMode(ctx context.Context) error {
 		return items
 	}
 
+	// getExtensionItems re-collects the loaded extension list, used by the
+	// TUI after an extension hot-reload to refresh the [Extensions] row.
+	getExtensionItems := func() []ui.ExtensionItem {
+		cwd, _ := os.Getwd()
+		return buildExtensionItems(kitInstance, cwd)
+	}
+
 	// Build extension UI providers once (shared between both modes).
 	getWidgets := widgetProviderForUI(kitInstance)
 	getHeader := headerProviderForUI(kitInstance)
@@ -1165,7 +1203,10 @@ func runNormalMode(ctx context.Context) error {
 		extWatcher, watchErr := extensions.NewWatcher(watchDirs, func() {
 			if err := reloadExtensionsForUI(); err != nil {
 				log.Printf("auto-reload extensions failed: %v", err)
+				appInstance.PrintFromExtension("error", fmt.Sprintf("Extension auto-reload failed: %v", err))
+				return
 			}
+			appInstance.PrintFromExtension("info", "Extensions reloaded.")
 		})
 		if watchErr != nil {
 			log.Printf("extension file watcher not started: %v", watchErr)
@@ -1184,6 +1225,7 @@ func runNormalMode(ctx context.Context) error {
 		promptDirs := watcher.CollectDirs(
 			[]string{
 				filepath.Join(homeDir, ".kit", "prompts"),
+				prompts.GlobalDir(),
 				filepath.Join(cwd, ".kit", "prompts"),
 			},
 			append(promptTemplatePaths, viper.GetStringSlice("prompts")...),
@@ -1222,7 +1264,7 @@ func runNormalMode(ctx context.Context) error {
 
 	// Check if running in non-interactive mode
 	if positionalPrompt != "" {
-		return runNonInteractiveModeApp(ctx, appInstance, cli, positionalPrompt, quietFlag, jsonFlag, noExitFlag, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI)
+		return runNonInteractiveModeApp(ctx, appInstance, cli, positionalPrompt, quietFlag, jsonFlag, noExitFlag, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI)
 	}
 
 	// Quiet mode is not allowed in interactive mode
@@ -1230,7 +1272,7 @@ func runNormalMode(ctx context.Context) error {
 		return fmt.Errorf("--quiet requires a prompt")
 	}
 
-	return runInteractiveModeBubbleTea(ctx, appInstance, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI, startupExtensionMessages)
+	return runInteractiveModeBubbleTea(ctx, appInstance, modelName, parsedProvider, kitInstance.GetLoadingMessage(), serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModelForUI, emitModelChangeForUI, kitInstance.IsReasoningModel(), kitInstance.GetThinkingLevel(), setThinkingLevelForUI, switchSessionForUI, reloadExtensionsForUI, startupExtensionMessages)
 }
 
 // runNonInteractiveModeApp executes a single prompt via the app layer and exits,
@@ -1243,7 +1285,7 @@ func runNormalMode(ctx context.Context) error {
 //
 // When --no-exit is set, after the prompt completes the interactive BubbleTea
 // TUI is started so the user can continue the conversation.
-func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui.CLI, prompt string, quiet, jsonOutput, noExit bool, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error) error {
+func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui.CLI, prompt string, quiet, jsonOutput, noExit bool, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, extensionItems []ui.ExtensionItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getExtensionItems func() []ui.ExtensionItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error) error {
 	// Expand @file references in the prompt before sending to the agent.
 	// Text files are XML-inlined; binary files are extracted as multimodal parts.
 	var fileParts []kit.LLMFilePart
@@ -1304,7 +1346,7 @@ func runNonInteractiveModeApp(ctx context.Context, appInstance *app.App, cli *ui
 
 	// If --no-exit was requested, hand off to the interactive TUI.
 	if noExit {
-		return runInteractiveModeBubbleTea(ctx, appInstance, modelName, providerName, loadingMessage, serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, getPromptTemplates, getSkillItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel, switchSession, reloadExtensions, nil)
+		return runInteractiveModeBubbleTea(ctx, appInstance, modelName, providerName, loadingMessage, serverNames, toolNames, mcpToolCount, extensionToolCount, usageTracker, extCommands, promptTemplates, contextPaths, skillItems, extensionItems, getPromptTemplates, getSkillItems, getExtensionItems, getToolNames, getMCPToolCount, mcpPrompts, getMCPPrompts, expandMCPPrompt, getWidgets, getHeader, getFooter, getToolRenderer, getEditorInterceptor, getUIVisibility, getStatusBarEntries, emitBeforeFork, emitBeforeSessionSwitch, getGlobalShortcuts, getExtensionCommands, setModel, emitModelChange, isReasoningModel, thinkingLevel, setThinkingLevel, switchSession, reloadExtensions, nil)
 	}
 
 	return nil
@@ -1402,7 +1444,7 @@ func writeJSONError(err error) {
 //  4. Calls program.Run() which blocks until the user quits (Ctrl+C or /quit).
 //
 // SetupCLI is not used for interactive mode; the TUI (AppModel) handles its own rendering.
-func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error, startupExtensionMessages []string) error {
+func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelName, providerName, loadingMessage string, serverNames, toolNames []string, mcpToolCount, extensionToolCount int, usageTracker *ui.UsageTracker, extCommands []commands.ExtensionCommand, promptTemplates []*prompts.PromptTemplate, contextPaths []string, skillItems []ui.SkillItem, extensionItems []ui.ExtensionItem, getPromptTemplates func() []*prompts.PromptTemplate, getSkillItems func() []ui.SkillItem, getExtensionItems func() []ui.ExtensionItem, getToolNames func() []string, getMCPToolCount func() int, mcpPrompts []ui.MCPPromptInfo, getMCPPrompts func() []ui.MCPPromptInfo, expandMCPPrompt func(string, string, map[string]string) (*ui.MCPPromptExpandResult, error), getWidgets func(string) []ui.WidgetData, getHeader, getFooter func() *ui.WidgetData, getToolRenderer func(string) *ui.ToolRendererData, getEditorInterceptor func() *ui.EditorInterceptor, getUIVisibility func() *ui.UIVisibility, getStatusBarEntries func() []ui.StatusBarEntryData, emitBeforeFork func(string, bool, string) (bool, string), emitBeforeSessionSwitch func(string) (bool, string), getGlobalShortcuts func() map[string]func(), getExtensionCommands func() []commands.ExtensionCommand, setModel func(string) error, emitModelChange func(string, string, string), isReasoningModel bool, thinkingLevel string, setThinkingLevel func(string) error, switchSession func(string) error, reloadExtensions func() error, startupExtensionMessages []string) error {
 	// Redirect all log output (stdlib and charm) to a file so that log
 	// messages don't write to stderr and corrupt the TUI. Bubble Tea
 	// captures stdout for rendering; any stray stderr output from
@@ -1447,6 +1489,8 @@ func runInteractiveModeBubbleTea(_ context.Context, appInstance *app.App, modelN
 		ContextPaths:             contextPaths,
 		SkillItems:               skillItems,
 		GetSkillItems:            getSkillItems,
+		ExtensionItems:           extensionItems,
+		GetExtensionItems:        getExtensionItems,
 		StartupExtensionMessages: startupExtensionMessages,
 		GetWidgets:               getWidgets,
 		GetHeader:                getHeader,

examples/extensions/subagent-monitor_test.go

@@ -13,7 +13,7 @@ import (
 // without panicking and properly guards nil ctx calls.
 func TestSubagentMonitor_SessionStart(t *testing.T) {
 	harness := test.New(t)
-	harness.LoadFile("../../.kit/extensions/subagent-monitor.go")
+	harness.LoadFile("./subagent-monitor.go")
 
 	// Emit SessionStart - should not panic even with nil ctx functions
 	_, err := harness.Emit(extensions.SessionStartEvent{SessionID: "test-session"})
@@ -26,7 +26,7 @@ func TestSubagentMonitor_SessionStart(t *testing.T) {
 // creates entries and emits widget updates.
 func TestSubagentMonitor_SubagentLifecycle(t *testing.T) {
 	harness := test.New(t)
-	harness.LoadFile("../../.kit/extensions/subagent-monitor.go")
+	harness.LoadFile("./subagent-monitor.go")
 
 	// Start session
 	_, err := harness.Emit(extensions.SessionStartEvent{SessionID: "test-session"})
@@ -84,7 +84,7 @@ func TestSubagentMonitor_SubagentLifecycle(t *testing.T) {
 // TestSubagentMonitor_MultipleSubagents verifies multiple parallel subagents.
 func TestSubagentMonitor_MultipleSubagents(t *testing.T) {
 	harness := test.New(t)
-	harness.LoadFile("../../.kit/extensions/subagent-monitor.go")
+	harness.LoadFile("./subagent-monitor.go")
 
 	_, err := harness.Emit(extensions.SessionStartEvent{SessionID: "test-session"})
 	if err != nil {
@@ -134,7 +134,7 @@ func TestSubagentMonitor_MultipleSubagents(t *testing.T) {
 // subagents emit events concurrently from different goroutines.
 func TestSubagentMonitor_ConcurrentSubagents(t *testing.T) {
 	harness := test.New(t)
-	harness.LoadFile("../../.kit/extensions/subagent-monitor.go")
+	harness.LoadFile("./subagent-monitor.go")
 
 	_, err := harness.Emit(extensions.SessionStartEvent{SessionID: "test-session"})
 	if err != nil {
@@ -186,7 +186,7 @@ func TestSubagentMonitor_ConcurrentSubagents(t *testing.T) {
 // even with nil ctx functions.
 func TestSubagentMonitor_SessionShutdown(t *testing.T) {
 	harness := test.New(t)
-	harness.LoadFile("../../.kit/extensions/subagent-monitor.go")
+	harness.LoadFile("./subagent-monitor.go")
 
 	// Start then shutdown
 	_, err := harness.Emit(extensions.SessionStartEvent{SessionID: "test-session"})

examples/sdk/README.md

@@ -42,4 +42,14 @@ defer host.Close()
 response, err := host.Prompt(ctx, "Hello!")
 ```
 
+Or use the functional-options constructor for quick setups (streaming defaults on):
+
+```go
+host, err := kit.NewAgent(ctx,
+    kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+    kit.WithSystemPrompt("You are a helpful assistant."),
+    kit.Ephemeral(),
+)
+```
+
 See the [SDK README](../../pkg/kit/README.md) for the full API reference.

go.mod

@@ -1,32 +1,32 @@
 module github.com/mark3labs/kit
 
-go 1.26.2
+go 1.26.3
 
 require (
 	charm.land/bubbles/v2 v2.1.0
 	charm.land/bubbletea/v2 v2.0.6
-	charm.land/fantasy v0.23.0
+	charm.land/fantasy v0.25.0
 	charm.land/huh/v2 v2.0.3
 	charm.land/lipgloss/v2 v2.0.3
-	github.com/alecthomas/chroma/v2 v2.24.1
+	github.com/alecthomas/chroma/v2 v2.26.1
 	github.com/atotto/clipboard v0.1.4
 	github.com/aymanbagabas/go-udiff v0.4.1
 	github.com/charmbracelet/fang v1.0.0
 	github.com/charmbracelet/log v1.0.0
 	github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266
-	github.com/charmbracelet/ultraviolet v0.0.0-20260428153724-66037269d7be
+	github.com/charmbracelet/ultraviolet v0.0.0-20260525132238-948f4557a654
 	github.com/charmbracelet/x/editor v0.2.0
 	github.com/clipperhouse/displaywidth v0.11.0
 	github.com/clipperhouse/uax29/v2 v2.7.0
-	github.com/coder/acp-go-sdk v0.12.2
+	github.com/coder/acp-go-sdk v0.13.0
 	github.com/fsnotify/fsnotify v1.10.1
 	github.com/indaco/herald v0.13.0
 	github.com/indaco/herald-md v0.3.0
-	github.com/mark3labs/mcp-go v0.51.0
+	github.com/mark3labs/mcp-go v0.54.1
 	github.com/spf13/cobra v1.10.2
 	github.com/spf13/viper v1.21.0
 	github.com/traefik/yaegi v0.16.1
-	golang.org/x/term v0.42.0
+	golang.org/x/term v0.43.0
 	gopkg.in/yaml.v3 v3.0.1
 )
 
@@ -37,21 +37,21 @@ require (
 	cloud.google.com/go/compute/metadata v0.9.0 // indirect
 	github.com/Azure/azure-sdk-for-go/sdk/azcore v1.21.1 // indirect
 	github.com/Azure/azure-sdk-for-go/sdk/internal v1.12.0 // indirect
-	github.com/aws/aws-sdk-go-v2 v1.41.7 // indirect
+	github.com/aws/aws-sdk-go-v2 v1.41.8 // indirect
 	github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.10 // indirect
-	github.com/aws/aws-sdk-go-v2/config v1.32.17 // indirect
-	github.com/aws/aws-sdk-go-v2/credentials v1.19.16 // indirect
-	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.23 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.23 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.23 // indirect
-	github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.24 // indirect
+	github.com/aws/aws-sdk-go-v2/config v1.32.19 // indirect
+	github.com/aws/aws-sdk-go-v2/credentials v1.19.18 // indirect
+	github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.24 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.24 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.24 // indirect
+	github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.25 // indirect
 	github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.9 // indirect
-	github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.23 // indirect
-	github.com/aws/aws-sdk-go-v2/service/signin v1.0.11 // indirect
-	github.com/aws/aws-sdk-go-v2/service/sso v1.30.17 // indirect
-	github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.21 // indirect
-	github.com/aws/aws-sdk-go-v2/service/sts v1.42.1 // indirect
-	github.com/aws/smithy-go v1.25.1 // indirect
+	github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.24 // indirect
+	github.com/aws/aws-sdk-go-v2/service/signin v1.1.0 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sso v1.30.18 // indirect
+	github.com/aws/aws-sdk-go-v2/service/ssooidc v1.36.1 // indirect
+	github.com/aws/aws-sdk-go-v2/service/sts v1.42.2 // indirect
+	github.com/aws/smithy-go v1.26.0 // indirect
 	github.com/catppuccin/go v0.3.0 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/charmbracelet/anthropic-sdk-go v0.0.0-20260223140439-63879b0b8dab // indirect
@@ -59,17 +59,18 @@ require (
 	github.com/charmbracelet/harmonica v0.2.0 // indirect
 	github.com/charmbracelet/lipgloss v1.1.1-0.20250404203927-76690c660834 // indirect
 	github.com/charmbracelet/x/cellbuf v0.0.15 // indirect
-	github.com/charmbracelet/x/exp/charmtone v0.0.0-20260503005035-c113ba3d2310 // indirect
+	github.com/charmbracelet/x/exp/charmtone v0.0.0-20260527151214-009e6338d40d // indirect
 	github.com/charmbracelet/x/exp/ordered v0.1.0 // indirect
-	github.com/charmbracelet/x/exp/slice v0.0.0-20260503005035-c113ba3d2310 // indirect
+	github.com/charmbracelet/x/exp/slice v0.0.0-20260527151214-009e6338d40d // indirect
 	github.com/charmbracelet/x/exp/strings v0.1.0 // indirect
 	github.com/charmbracelet/x/json v0.2.0 // indirect
 	github.com/charmbracelet/x/termios v0.1.1 // indirect
 	github.com/charmbracelet/x/windows v0.2.2 // indirect
 	github.com/dlclark/regexp2 v1.12.0 // indirect
+	github.com/dlclark/regexp2/v2 v2.1.1 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
-	github.com/go-json-experiment/json v0.0.0-20260430182902-b6187a392ed4 // indirect
+	github.com/go-json-experiment/json v0.0.0-20260520185125-572e7c383686 // indirect
 	github.com/go-logfmt/logfmt v0.6.1 // indirect
 	github.com/go-logr/logr v1.4.3 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
@@ -79,13 +80,13 @@ require (
 	github.com/google/jsonschema-go v0.4.3 // indirect
 	github.com/google/s2a-go v0.1.9 // indirect
 	github.com/google/uuid v1.6.0 // indirect
-	github.com/googleapis/enterprise-certificate-proxy v0.3.15 // indirect
+	github.com/googleapis/enterprise-certificate-proxy v0.3.16 // indirect
 	github.com/googleapis/gax-go/v2 v2.22.0 // indirect
 	github.com/gorilla/websocket v1.5.3 // indirect
-	github.com/kaptinlin/go-i18n v0.4.7 // indirect
-	github.com/kaptinlin/jsonpointer v0.4.21 // indirect
+	github.com/kaptinlin/go-i18n v0.4.5 // indirect
+	github.com/kaptinlin/jsonpointer v0.4.25 // indirect
 	github.com/kaptinlin/jsonschema v0.7.13 // indirect
-	github.com/kaptinlin/messageformat-go v0.6.3 // indirect
+	github.com/kaptinlin/messageformat-go v0.6.0 // indirect
 	github.com/mitchellh/hashstructure/v2 v2.0.2 // indirect
 	github.com/muesli/mango v0.2.0 // indirect
 	github.com/muesli/mango-cobra v1.3.0 // indirect
@@ -97,7 +98,7 @@ require (
 	github.com/spf13/afero v1.15.0 // indirect
 	github.com/spf13/cast v1.10.0 // indirect
 	github.com/subosito/gotenv v1.6.0 // indirect
-	github.com/tidwall/gjson v1.18.0 // indirect
+	github.com/tidwall/gjson v1.19.0 // indirect
 	github.com/tidwall/match v1.2.0 // indirect
 	github.com/tidwall/pretty v1.2.1 // indirect
 	github.com/tidwall/sjson v1.2.5 // indirect
@@ -105,21 +106,21 @@ require (
 	github.com/yosida95/uritemplate/v3 v3.0.2 // indirect
 	github.com/yuin/goldmark v1.8.2 // indirect
 	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
-	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.68.0 // indirect
-	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0 // indirect
-	go.opentelemetry.io/otel v1.43.0 // indirect
-	go.opentelemetry.io/otel/metric v1.43.0 // indirect
-	go.opentelemetry.io/otel/trace v1.43.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.69.0 // indirect
+	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.69.0 // indirect
+	go.opentelemetry.io/otel v1.44.0 // indirect
+	go.opentelemetry.io/otel/metric v1.44.0 // indirect
+	go.opentelemetry.io/otel/trace v1.44.0 // indirect
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
-	golang.org/x/crypto v0.50.0 // indirect
-	golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f // indirect
-	golang.org/x/net v0.53.0 // indirect
+	golang.org/x/crypto v0.52.0 // indirect
+	golang.org/x/exp v0.0.0-20260528193900-50dc527dd6c7 // indirect
+	golang.org/x/net v0.55.0 // indirect
 	golang.org/x/oauth2 v0.36.0 // indirect
 	golang.org/x/time v0.15.0 // indirect
-	google.golang.org/api v0.277.0 // indirect
-	google.golang.org/genai v1.55.0 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20260427160629-7cedc36a6bc4 // indirect
-	google.golang.org/grpc v1.81.0 // indirect
+	google.golang.org/api v0.282.0 // indirect
+	google.golang.org/genai v1.58.0 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20260526163538-3dc84a4a5aaa // indirect
+	google.golang.org/grpc v1.81.1 // indirect
 	google.golang.org/protobuf v1.36.11 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 )
@@ -137,6 +138,6 @@ require (
 	github.com/rivo/uniseg v0.4.7 // indirect
 	github.com/spf13/pflag v1.0.10
 	golang.org/x/sync v0.20.0 // indirect
-	golang.org/x/sys v0.43.0 // indirect
-	golang.org/x/text v0.36.0
+	golang.org/x/sys v0.45.0 // indirect
+	golang.org/x/text v0.37.0
 )

go.sum

@@ -2,8 +2,8 @@ charm.land/bubbles/v2 v2.1.0 h1:YSnNh5cPYlYjPxRrzs5VEn3vwhtEn3jVGRBT3M7/I0g=
 charm.land/bubbles/v2 v2.1.0/go.mod h1:l97h4hym2hvWBVfmJDtrEHHCtkIKeTEb3TTJ4ZOB3wY=
 charm.land/bubbletea/v2 v2.0.6 h1:UHN/91OyuhaOFGSrBXQ/hMZD8IO1Uc4BvHlgHXL2WJo=
 charm.land/bubbletea/v2 v2.0.6/go.mod h1:MH/D8ZLlN3op37vQvijKuU29g3rqTp+aQapURFonF9g=
-charm.land/fantasy v0.23.0 h1:pocjwC5CxfEg1Bpwb0raML2d5ijo3op33Mmd6hYJyo4=
-charm.land/fantasy v0.23.0/go.mod h1:4yzSsd9XmFEVjRnF1P0LTEbLTmQX6OLnPkrHaf7iruo=
+charm.land/fantasy v0.25.0 h1:oXOWY1ivmTSnhYGzAolscF8zKtavWZyBWv0LHRSwN5Q=
+charm.land/fantasy v0.25.0/go.mod h1:8QrWUzIcKwZQP+aAnC9vLu3iID6hu9/Jt+rPMiieBkc=
 charm.land/huh/v2 v2.0.3 h1:2cJsMqEPwSywGHvdlKsJyQKPtSJLVnFKyFbsYZTlLkU=
 charm.land/huh/v2 v2.0.3/go.mod h1:93eEveeeqn47MwiC3tf+2atZ2l7Is88rAtmZNZ8x9Wc=
 charm.land/lipgloss/v2 v2.0.3 h1:yM2zJ4Cf5Y51b7RHIwioil4ApI/aypFXXVHSwlM6RzU=
@@ -28,42 +28,42 @@ github.com/MakeNowJust/heredoc v1.0.0 h1:cXCdzVdstXyiTqTvfqk9SDHpKNjxuom+DOlyEeQ
 github.com/MakeNowJust/heredoc v1.0.0/go.mod h1:mG5amYoWBHf8vpLOuehzbGGw0EHxpZZ6lCpQ4fNJ8LE=
 github.com/alecthomas/assert/v2 v2.11.0 h1:2Q9r3ki8+JYXvGsDyBXwH3LcJ+WK5D0gc5E8vS6K3D0=
 github.com/alecthomas/assert/v2 v2.11.0/go.mod h1:Bze95FyfUr7x34QZrjL+XP+0qgp/zg8yS+TtBj1WA3k=
-github.com/alecthomas/chroma/v2 v2.24.1 h1:m5ffpfZbIb++k8AqFEKy9uVgY12xIQtBsQlc6DfZJQM=
-github.com/alecthomas/chroma/v2 v2.24.1/go.mod h1:l+ohZ9xRXIbGe7cIW+YZgOGbvuVLjMps/FYN/CwuabI=
+github.com/alecthomas/chroma/v2 v2.26.1 h1:2X21EdxGZNv5GF9mG5u+uzc02GCFyGxbcBm3Grd9A78=
+github.com/alecthomas/chroma/v2 v2.26.1/go.mod h1:lxhRRa9H4hPmRLOOdYga4zkQIQjq3dtrrdwQeCfu78Y=
 github.com/alecthomas/repr v0.5.2 h1:SU73FTI9D1P5UNtvseffFSGmdNci/O6RsqzeXJtP0Qs=
 github.com/alecthomas/repr v0.5.2/go.mod h1:Fr0507jx4eOXV7AlPV6AVZLYrLIuIeSOWtW57eE/O/4=
 github.com/atotto/clipboard v0.1.4 h1:EH0zSVneZPSuFR11BlR9YppQTVDbh5+16AmcJi4g1z4=
 github.com/atotto/clipboard v0.1.4/go.mod h1:ZY9tmq7sm5xIbd9bOK4onWV4S6X0u6GY7Vn0Yu86PYI=
-github.com/aws/aws-sdk-go-v2 v1.41.7 h1:DWpAJt66FmnnaRIOT/8ASTucrvuDPZASqhhLey6tLY8=
-github.com/aws/aws-sdk-go-v2 v1.41.7/go.mod h1:4LAfZOPHNVNQEckOACQx60Y8pSRjIkNZQz1w92xpMJc=
+github.com/aws/aws-sdk-go-v2 v1.41.8 h1:sRs7nG6/RiEBZ/K5UO2sNw0w40U02Nmz1VtARloTZXk=
+github.com/aws/aws-sdk-go-v2 v1.41.8/go.mod h1:4LAfZOPHNVNQEckOACQx60Y8pSRjIkNZQz1w92xpMJc=
 github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.10 h1:gx1AwW1Iyk9Z9dD9F4akX5gnN3QZwUB20GGKH/I+Rho=
 github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.10/go.mod h1:qqY157uZoqm5OXq/amuaBJyC9hgBCBQnsaWnPe905GY=
-github.com/aws/aws-sdk-go-v2/config v1.32.17 h1:FpL4/758/diKwqbytU0prpuiu60fgXKUWCpDJtApclU=
-github.com/aws/aws-sdk-go-v2/config v1.32.17/go.mod h1:OXqUMzgXytfoF9JaKkhrOYsyh72t9G+MJH8mMRaexOE=
-github.com/aws/aws-sdk-go-v2/credentials v1.19.16 h1:r3RJBuU7X9ibt8RHbMjWE6y60QbKBiII6wSrXnapxSU=
-github.com/aws/aws-sdk-go-v2/credentials v1.19.16/go.mod h1:6cx7zqDENJDbBIIWX6P8s0h6hqHC8Avbjh9Dseo27ug=
-github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.23 h1:UuSfcORqNSz/ey3VPRS8TcVH2Ikf0/sC+Hdj400QI6U=
-github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.23/go.mod h1:+G/OSGiOFnSOkYloKj/9M35s74LgVAdJBSD5lsFfqKg=
-github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.23 h1:GpT/TrnBYuE5gan2cZbTtvP+JlHsutdmlV2YfEyNde0=
-github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.23/go.mod h1:xYWD6BS9ywC5bS3sz9Xh04whO/hzK2plt2Zkyrp4JuA=
-github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.23 h1:bpd8vxhlQi2r1hiueOw02f/duEPTMK59Q4QMAoTTtTo=
-github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.23/go.mod h1:15DfR2nw+CRHIk0tqNyifu3G1YdAOy68RftkhMDDwYk=
-github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.24 h1:OQqn11BtaYv1WLUowvcA30MpzIu8Ti4pcLPIIyoKZrA=
-github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.24/go.mod h1:X5ZJyfwVrWA96GzPmUCWFQaEARPR7gCrpq2E92PJwAE=
+github.com/aws/aws-sdk-go-v2/config v1.32.19 h1:qRhIJMbevHUvIE7X4TK8N8zye5+5AhapcslPrvB+qKE=
+github.com/aws/aws-sdk-go-v2/config v1.32.19/go.mod h1:RbJ24nfoya63+Mf5VI+CGCGk9vEdv28xPeii+gojRYs=
+github.com/aws/aws-sdk-go-v2/credentials v1.19.18 h1:GcXQz2M/0ZvMo0v5DakUqbDBeBM1ZNaivkolEF4Esgw=
+github.com/aws/aws-sdk-go-v2/credentials v1.19.18/go.mod h1:sHJ06tMGcD3ZpmMyJqV+VBsGilhSIZPIN+ZFy5Dg0C4=
+github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.24 h1:FQm5ApnyzkuJdXLGskPce83CK1CQKC4RUnIHKVe4BU4=
+github.com/aws/aws-sdk-go-v2/feature/ec2/imds v1.18.24/go.mod h1:JsC7dqQc55MlZ5mvNsDMMge71u8pVcSzU3RNz2h/5yQ=
+github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.24 h1:u6kJU2i0va1AgtJsH3RdWKWqHULlTh7zHwb35Womf74=
+github.com/aws/aws-sdk-go-v2/internal/configsources v1.4.24/go.mod h1:7GY+xLcXOFUpCkNwDReft9qOAVg54A4/AnjHIU7sSAY=
+github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.24 h1:Xhbcf3KugX6vX7SDyUK205Oicyfg7EGuvoVNyP5L6DM=
+github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.7.24/go.mod h1:rwDgb2HNOGZsnTHylOUedM7Vnl+bCfnXDqUNPsFWYfk=
+github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.25 h1:54CTMmlJ71Rk2dYvM9qZOob+39wjlVja2zDLxCu69Ew=
+github.com/aws/aws-sdk-go-v2/internal/v4a v1.4.25/go.mod h1:BZaHqxsS9vN1fvV5EfEl0OBLOk5+AajWsMu6MjqnZB4=
 github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.9 h1:FLudkZLt5ci0ozzgkVo8BJGwvqNaZbTWb3UcucAateA=
 github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.13.9/go.mod h1:w7wZ/s9qK7c8g4al+UyoF1Sp/Z45UwMGcqIzLWVQHWk=
-github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.23 h1:pbrxO/kuIwgEsOPLkaHu0O+m4fNgLU8B3vxQ+72jTPw=
-github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.23/go.mod h1:/CMNUqoj46HpS3MNRDEDIwcgEnrtZlKRaHNaHxIFpNA=
-github.com/aws/aws-sdk-go-v2/service/signin v1.0.11 h1:TdJ+HdzOBhU8+iVAOGUTU63VXopcumCOF1paFulHWZc=
-github.com/aws/aws-sdk-go-v2/service/signin v1.0.11/go.mod h1:R82ZRExE/nheo0N+T8zHPcLRTcH8MGsnR3BiVGX0TwI=
-github.com/aws/aws-sdk-go-v2/service/sso v1.30.17 h1:7byT8HUWrgoRp6sXjxtZwgOKfhss5fW6SkLBtqzgRoE=
-github.com/aws/aws-sdk-go-v2/service/sso v1.30.17/go.mod h1:xNWknVi4Ezm1vg1QsB/5EWpAJURq22uqd38U8qKvOJc=
-github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.21 h1:+1Kl1zx6bWi4X7cKi3VYh29h8BvsCoHQEQ6ST9X8w7w=
-github.com/aws/aws-sdk-go-v2/service/ssooidc v1.35.21/go.mod h1:4vIRDq+CJB2xFAXZ+YgGUTiEft7oAQlhIs71xcSeuVg=
-github.com/aws/aws-sdk-go-v2/service/sts v1.42.1 h1:F/M5Y9I3nwr2IEpshZgh1GeHpOItExNM9L1euNuh/fk=
-github.com/aws/aws-sdk-go-v2/service/sts v1.42.1/go.mod h1:mTNxImtovCOEEuD65mKW7DCsL+2gjEH+RPEAexAzAio=
-github.com/aws/smithy-go v1.25.1 h1:J8ERsGSU7d+aCmdQur5Txg6bVoYelvQJgtZehD12GkI=
-github.com/aws/smithy-go v1.25.1/go.mod h1:YE2RhdIuDbA5E5bTdciG9KrW3+TiEONeUWCqxX9i1Fc=
+github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.24 h1:CQW2FTrflfoslYWLf3fv7vG28Q219+v8YJS5QTQb2+Y=
+github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.13.24/go.mod h1:Xfx13T+u3nH6EEzgl9fBSO6nDRmze1FvnZNYkctQ2zw=
+github.com/aws/aws-sdk-go-v2/service/signin v1.1.0 h1:yQo3eZ5qFaL1sJWqs1nL6j3yPHA2/R7c6tQ4T+0IO10=
+github.com/aws/aws-sdk-go-v2/service/signin v1.1.0/go.mod h1:3Zzou41Qt/ueXfIzHvTEjDNuR5IjCUBVF01SNhrt1e8=
+github.com/aws/aws-sdk-go-v2/service/sso v1.30.18 h1:ApLTFdAZfDhZSiY5uskwECKHkSNNF83y2Ru2r7SezWA=
+github.com/aws/aws-sdk-go-v2/service/sso v1.30.18/go.mod h1:A9K9qx2l6nK89hp+a350FdGfRkrkH5HdiEjHbiy/Q/c=
+github.com/aws/aws-sdk-go-v2/service/ssooidc v1.36.1 h1:4VD7TIZOGzehrgQ8vDE+1c6BQW4ErZPGY8ohZT5LXEE=
+github.com/aws/aws-sdk-go-v2/service/ssooidc v1.36.1/go.mod h1:er0SFJfdV89Rit5hIJu/EXtv+qC2XMnxoksLmcUFkqM=
+github.com/aws/aws-sdk-go-v2/service/sts v1.42.2 h1:XKnxlM4KZH1gktcsh3zSWc7GW4KivEv/OkifmHOhCUY=
+github.com/aws/aws-sdk-go-v2/service/sts v1.42.2/go.mod h1:KJYmkQaFB3SUW2j3aBkPsxNmAb4ZsSOvbvCpuxzHJA0=
+github.com/aws/smithy-go v1.26.0 h1:9ouqbi+NyKP7fV3Te7UElCwdAb6Y8uk7LGwPE5tVe/s=
+github.com/aws/smithy-go v1.26.0/go.mod h1:YE2RhdIuDbA5E5bTdciG9KrW3+TiEONeUWCqxX9i1Fc=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1 h1:HwpRHbFMcZLEVr42D4p7XBqjyuxQH5SMiErDT4WkJ2k=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1/go.mod h1:uYgXzlJ7ZpABp8OJ+exZzJJhRNQ2ASbcXHWsFqH8hp8=
 github.com/aymanbagabas/go-udiff v0.4.1 h1:OEIrQ8maEeDBXQDoGCbbTTXYJMYRCRO1fnodZ12Gv5o=
@@ -86,8 +86,8 @@ github.com/charmbracelet/log v1.0.0 h1:HVVVMmfOorfj3BA9i8X8UL69Hoz9lI0PYwXfJvOdR
 github.com/charmbracelet/log v1.0.0/go.mod h1:uYgY3SmLpwJWxmlrPwXvzVYujxis1vAKRV/0VQB7yWA=
 github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266 h1:BW/sZtyd1JyYy0h5adMm3tzpNyL857LWjuTRET6OhpY=
 github.com/charmbracelet/openai-go v0.0.0-20260319145158-d0740cc34266/go.mod h1:1DahUaExbUZx/jD+FNT2PKP4L9rLE5+ZBRuI8mZjd/E=
-github.com/charmbracelet/ultraviolet v0.0.0-20260428153724-66037269d7be h1:j7w8VP/D4lu5+/4GamMmFy8nrtadcl82/fjvDgSHwLo=
-github.com/charmbracelet/ultraviolet v0.0.0-20260428153724-66037269d7be/go.mod h1:3YdTxlnV/L0bQ3VN8WOSw8doF7LZV/xawUQ4MuAPDvo=
+github.com/charmbracelet/ultraviolet v0.0.0-20260525132238-948f4557a654 h1:FpSYhY28ucg9ZRr+2wj67FAQ0Ey5yiK0072PmRDJNek=
+github.com/charmbracelet/ultraviolet v0.0.0-20260525132238-948f4557a654/go.mod h1:hFpumms29Smx3LStRfku8vcCTBe1Kq8aCXtHUJa3mjY=
 github.com/charmbracelet/x/ansi v0.11.7 h1:kzv1kJvjg2S3r9KHo8hDdHFQLEqn4RBCb39dAYC84jI=
 github.com/charmbracelet/x/ansi v0.11.7/go.mod h1:9qGpnAVYz+8ACONkZBUWPtL7lulP9No6p1epAihUZwQ=
 github.com/charmbracelet/x/cellbuf v0.0.15 h1:ur3pZy0o6z/R7EylET877CBxaiE1Sp1GMxoFPAIztPI=
@@ -98,14 +98,14 @@ github.com/charmbracelet/x/editor v0.2.0 h1:7XLUKtaRaB8jN7bWU2p2UChiySyaAuIfYiIR
 github.com/charmbracelet/x/editor v0.2.0/go.mod h1:p3oQ28TSL3YPd+GKJ1fHWcp+7bVGpedHpXmo0D6t1dY=
 github.com/charmbracelet/x/errors v0.0.0-20240508181413-e8d8b6e2de86 h1:JSt3B+U9iqk37QUU2Rvb6DSBYRLtWqFqfxf8l5hOZUA=
 github.com/charmbracelet/x/errors v0.0.0-20240508181413-e8d8b6e2de86/go.mod h1:2P0UgXMEa6TsToMSuFqKFQR+fZTO9CNGUNokkPatT/0=
-github.com/charmbracelet/x/exp/charmtone v0.0.0-20260503005035-c113ba3d2310 h1:rByFKh9JgQScu7oy0+TlUbC2e93woW/QNZmNXbbbw/E=
-github.com/charmbracelet/x/exp/charmtone v0.0.0-20260503005035-c113ba3d2310/go.mod h1:nsExn0DGyX0lh9LwLHTn2Gg+hafdzfSXnC+QmEJTZFY=
+github.com/charmbracelet/x/exp/charmtone v0.0.0-20260527151214-009e6338d40d h1:sMilwx1YIYTrQva6jsB522AoRYAerNaDIKP4ZPtUq0A=
+github.com/charmbracelet/x/exp/charmtone v0.0.0-20260527151214-009e6338d40d/go.mod h1:nsExn0DGyX0lh9LwLHTn2Gg+hafdzfSXnC+QmEJTZFY=
 github.com/charmbracelet/x/exp/golden v0.0.0-20250806222409-83e3a29d542f h1:pk6gmGpCE7F3FcjaOEKYriCvpmIN4+6OS/RD0vm4uIA=
 github.com/charmbracelet/x/exp/golden v0.0.0-20250806222409-83e3a29d542f/go.mod h1:IfZAMTHB6XkZSeXUqriemErjAWCCzT0LwjKFYCZyw0I=
 github.com/charmbracelet/x/exp/ordered v0.1.0 h1:55/qLwjIh0gL0Vni+QAWk7T/qRVP6sBf+2agPBgnOFE=
 github.com/charmbracelet/x/exp/ordered v0.1.0/go.mod h1:5UHwmG+is5THxMyCJHNPCn2/ecI07aKNrW+LcResjJ8=
-github.com/charmbracelet/x/exp/slice v0.0.0-20260503005035-c113ba3d2310 h1:PMjHdSo8Vpq9psUw9BoHo9JLPMkm9Hqb+Whk64n3AQQ=
-github.com/charmbracelet/x/exp/slice v0.0.0-20260503005035-c113ba3d2310/go.mod h1:vqEfX6xzqW1pKKZUUiFOKg0OQ7bCh54Q2vR/tserrRA=
+github.com/charmbracelet/x/exp/slice v0.0.0-20260527151214-009e6338d40d h1:RxcAR+vJCoD8QqT1cqLtkQKw+1cqvjqnu5IpPqYzPco=
+github.com/charmbracelet/x/exp/slice v0.0.0-20260527151214-009e6338d40d/go.mod h1:vqEfX6xzqW1pKKZUUiFOKg0OQ7bCh54Q2vR/tserrRA=
 github.com/charmbracelet/x/exp/strings v0.1.0 h1:i69S2XI7uG1u4NLGeJPSYU++Nmjvpo9nwd6aoEm7gkA=
 github.com/charmbracelet/x/exp/strings v0.1.0/go.mod h1:/ehtMPNh9K4odGFkqYJKpIYyePhdp1hLBRvyY4bWkH8=
 github.com/charmbracelet/x/json v0.2.0 h1:DqB+ZGx2h+Z+1s98HOuOyli+i97wsFQIxP2ZQANTPrQ=
@@ -124,8 +124,8 @@ github.com/clipperhouse/uax29/v2 v2.7.0 h1:+gs4oBZ2gPfVrKPthwbMzWZDaAFPGYK72F0NJ
 github.com/clipperhouse/uax29/v2 v2.7.0/go.mod h1:EFJ2TJMRUaplDxHKj1qAEhCtQPW2tJSwu5BF98AuoVM=
 github.com/cncf/xds/go v0.0.0-20260202195803-dba9d589def2 h1:aBangftG7EVZoUb69Os8IaYg++6uMOdKK83QtkkvJik=
 github.com/cncf/xds/go v0.0.0-20260202195803-dba9d589def2/go.mod h1:qwXFYgsP6T7XnJtbKlf1HP8AjxZZyzxMmc+Lq5GjlU4=
-github.com/coder/acp-go-sdk v0.12.2 h1:fpRJ8Z5HMSr5cZ5IywzFlFZcIxZOsto+laNVu7XelFA=
-github.com/coder/acp-go-sdk v0.12.2/go.mod h1:yKzM/3R9uELp4+nBAwwtkS0aN1FOFjo11CNPy37yFko=
+github.com/coder/acp-go-sdk v0.13.0 h1:IAKBDIbe/iBfKAGikeIndzb8fowt4ioD+gCtSU4HwMA=
+github.com/coder/acp-go-sdk v0.13.0/go.mod h1:yKzM/3R9uELp4+nBAwwtkS0aN1FOFjo11CNPy37yFko=
 github.com/cpuguy83/go-md2man/v2 v2.0.6/go.mod h1:oOW0eioCTA6cOiMLiUPZOpcVxMig6NIQQ7OS05n1F4g=
 github.com/creack/pty v1.1.24 h1:bJrF4RRfyJnbTJqzRLHzcGaZK1NeM5kTC9jGgovnR1s=
 github.com/creack/pty v1.1.24/go.mod h1:08sCNb52WyoAwi2QDyzUCTgcvVFhUzewun7wtTfvcwE=
@@ -133,6 +133,8 @@ github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/dlclark/regexp2 v1.12.0 h1:0j4c5qQmnC6XOWNjP3PIXURXN2gWx76rd3KvgdPkCz8=
 github.com/dlclark/regexp2 v1.12.0/go.mod h1:DHkYz0B9wPfa6wondMfaivmHpzrQ3v9q8cnmRbL6yW8=
+github.com/dlclark/regexp2/v2 v2.1.1 h1:LCUGyd9Wf+r+VVOl8Ny38JTpWJcAsdVnCIuhhtthmKw=
+github.com/dlclark/regexp2/v2 v2.1.1/go.mod h1:avUrQvPaLz2DrFNHJF0taWAFFX2C1GMSSoeiqFjcBmU=
 github.com/dnaeon/go-vcr v1.2.0 h1:zHCHvJYTMh1N7xnV7zf1m1GPBF9Ad0Jk/whtQ1663qI=
 github.com/dnaeon/go-vcr v1.2.0/go.mod h1:R4UdLID7HZT3taECzJs4YgbbH6PIGXB6W/sc5OLb6RQ=
 github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY=
@@ -148,8 +150,8 @@ github.com/frankban/quicktest v1.14.6 h1:7Xjx+VpznH+oBnejlPUj8oUpdxnVs4f8XU8WnHk
 github.com/frankban/quicktest v1.14.6/go.mod h1:4ptaffx2x8+WTWXmUCuVU6aPUX1/Mz7zb5vbUoiM6w0=
 github.com/fsnotify/fsnotify v1.10.1 h1:b0/UzAf9yR5rhf3RPm9gf3ehBPpf0oZKIjtpKrx59Ho=
 github.com/fsnotify/fsnotify v1.10.1/go.mod h1:TLheqan6HD6GBK6PrDWyDPBaEV8LspOxvPSjC+bVfgo=
-github.com/go-json-experiment/json v0.0.0-20260430182902-b6187a392ed4 h1:2WmHkJINIjgXXYDGik8d3oJvFA3DAwPy00csDJ3vo+o=
-github.com/go-json-experiment/json v0.0.0-20260430182902-b6187a392ed4/go.mod h1:tphK2c80bpPhMOI4v6bIc2xWywPfbqi1Z06+RcrMkDg=
+github.com/go-json-experiment/json v0.0.0-20260520185125-572e7c383686 h1:NZBJxCpbHS1gzS6xAmyxbJznosZIIPk9IB42v62UvKA=
+github.com/go-json-experiment/json v0.0.0-20260520185125-572e7c383686/go.mod h1:tphK2c80bpPhMOI4v6bIc2xWywPfbqi1Z06+RcrMkDg=
 github.com/go-logfmt/logfmt v0.6.1 h1:4hvbpePJKnIzH1B+8OR/JPbTx37NktoI9LE2QZBBkvE=
 github.com/go-logfmt/logfmt v0.6.1/go.mod h1:EV2pOAQoZaT1ZXZbqDl5hrymndi4SY9ED9/z6CO0XAk=
 github.com/go-logr/logr v1.2.2/go.mod h1:jdQByPbusPIv2/zmleS9BjJVeZ6kBagPoEUsqbVz/1A=
@@ -173,8 +175,8 @@ github.com/google/s2a-go v0.1.9 h1:LGD7gtMgezd8a/Xak7mEWL0PjoTQFvpRudN895yqKW0=
 github.com/google/s2a-go v0.1.9/go.mod h1:YA0Ei2ZQL3acow2O62kdp9UlnvMmU7kA6Eutn0dXayM=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
-github.com/googleapis/enterprise-certificate-proxy v0.3.15 h1:xolVQTEXusUcAA5UgtyRLjelpFFHWlPQ4XfWGc7MBas=
-github.com/googleapis/enterprise-certificate-proxy v0.3.15/go.mod h1:vqVt9yG9480NtzREnTlmGSBmFrA+bzb0yl0TxoBQXOg=
+github.com/googleapis/enterprise-certificate-proxy v0.3.16 h1:F/VPrx0YPBdksZJQdCAp0WUsqnNmZpUZszzfYt0M5Dw=
+github.com/googleapis/enterprise-certificate-proxy v0.3.16/go.mod h1:9Yb0eAkH/Xqhvv3zbeKf/+wMJqCeocWc6KIhDvEAuYE=
 github.com/googleapis/gax-go/v2 v2.22.0 h1:PjIWBpgGIVKGoCXuiCoP64altEJCj3/Ei+kSU5vlZD4=
 github.com/googleapis/gax-go/v2 v2.22.0/go.mod h1:irWBbALSr0Sk3qlqb9SyJ1h68WjgeFuiOzI4Rqw5+aY=
 github.com/gorilla/websocket v1.5.3 h1:saDtZ6Pbx/0u+bgYQ3q96pZgCzfhKXGPqt7kZ72aNNg=
@@ -187,14 +189,14 @@ github.com/indaco/herald v0.13.0 h1:+xVG9Fx5NpuWhwku/9IlRL6I009NnX4VUGKvlZHTRxU=
 github.com/indaco/herald v0.13.0/go.mod h1:T5g1+XLYvpjouhzAGHnAHDCKizhESkoV6+QPZ3DhgWA=
 github.com/indaco/herald-md v0.3.0 h1:hN1cKyrexPPM9PeHBsKuaWvIizSi/iYvM9yzRgtdb8M=
 github.com/indaco/herald-md v0.3.0/go.mod h1:RUHVaDSG45ymJjKyxpDwBocLXrZo93FB4OeYMsw9B9s=
-github.com/kaptinlin/go-i18n v0.4.7 h1:apjIIZHnGRyrkiX3vHj07F1BF6D0JLmV+VGSr1781Jc=
-github.com/kaptinlin/go-i18n v0.4.7/go.mod h1:+i1J0pFq/9i9ESC5qRMVkKwC+mdQTABhhBExpYOlbeM=
-github.com/kaptinlin/jsonpointer v0.4.21 h1:WVkwQbeerbHFcoXG7Yo/mlQhhZjWiTnagECEfwDXXa0=
-github.com/kaptinlin/jsonpointer v0.4.21/go.mod h1:Mo7+DX8RlQTFqS4dnYJl0izSP4ob+Rl5xO/mGDETgaU=
+github.com/kaptinlin/go-i18n v0.4.5 h1:9tIlo5A0RXth+yZJO2MG7Bhpu/X9PlzQnGz/qyYWNoY=
+github.com/kaptinlin/go-i18n v0.4.5/go.mod h1:mU/7BH4molY5lGZYBwBRKAaiJ70dWRHuqmQ0/pFLGno=
+github.com/kaptinlin/jsonpointer v0.4.25 h1:iJ197e8n+WwqaqBsa53FqG3rPJCg5oijyFXEXNWWC3E=
+github.com/kaptinlin/jsonpointer v0.4.25/go.mod h1:wVOBaXGGnP42YsMb6zev/3W5POTvspdNfh8DXzf8XS8=
 github.com/kaptinlin/jsonschema v0.7.13 h1:kahVXTy/rURL0XJjyQ9WELm59wEmXi6IY0TWswQEFvU=
 github.com/kaptinlin/jsonschema v0.7.13/go.mod h1:Uh0aUBusnhXDCEXJ2oimL/hx7YTo7F+sKniE+tM0ERc=
-github.com/kaptinlin/messageformat-go v0.6.3 h1:m9ZE/fCjnsk8bdkv7Qs56L/ZoHbmQqhz9mRZSAQLU5g=
-github.com/kaptinlin/messageformat-go v0.6.3/go.mod h1:2KOZ/hgo/SveZ+uyi7vPUpUXieX65Mppzbc3VpGyqKs=
+github.com/kaptinlin/messageformat-go v0.6.0 h1:D6jiXFsKW4/JG2CMddv/F6Rev9KVbCRKEzzV5QOAcpc=
+github.com/kaptinlin/messageformat-go v0.6.0/go.mod h1:NKjwS6e9u7DRhAK+vydjDDwJ7UbdHhYjk/yk2WPuZPs=
 github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
 github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
@@ -203,8 +205,8 @@ github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/lucasb-eyer/go-colorful v1.4.0 h1:UtrWVfLdarDgc44HcS7pYloGHJUjHV/4FwW4TvVgFr4=
 github.com/lucasb-eyer/go-colorful v1.4.0/go.mod h1:R4dSotOR9KMtayYi1e77YzuveK+i7ruzyGqttikkLy0=
-github.com/mark3labs/mcp-go v0.51.0 h1:e8AhEfxzcYt7XqYzwT7uzWNhnqpu3H1Tn7dEJB9Ygj8=
-github.com/mark3labs/mcp-go v0.51.0/go.mod h1:Zg9cB2HdwdMMVgY0xtTzq3KvYIOJQDsaut+jWjwDaQY=
+github.com/mark3labs/mcp-go v0.54.1 h1:Ap/ptEB9FtWzFKM8NDsTA7QDxerQOC06eZigrTldVj0=
+github.com/mark3labs/mcp-go v0.54.1/go.mod h1:+8WclSK1ZUweCP3hvktSji8n8ABG/95QaEkeVE/Uwas=
 github.com/mattn/go-isatty v0.0.22 h1:j8l17JJ9i6VGPUFUYoTUKPSgKe/83EYU2zBC7YNKMw4=
 github.com/mattn/go-isatty v0.0.22/go.mod h1:ZXfXG4SQHsB/w3ZeOYbR0PrPwLy+n6xiMrJlRFqopa4=
 github.com/mattn/go-runewidth v0.0.23 h1:7ykA0T0jkPpzSvMS5i9uoNn2Xy3R383f9HDx3RybWcw=
@@ -256,8 +258,8 @@ github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD
 github.com/subosito/gotenv v1.6.0 h1:9NlTDc1FTs4qu0DDq7AEtTPNw6SVm7uBMsUCUjABIf8=
 github.com/subosito/gotenv v1.6.0/go.mod h1:Dk4QP5c2W3ibzajGcXpNraDfq2IrhjMIvMSWPKKo0FU=
 github.com/tidwall/gjson v1.14.2/go.mod h1:/wbyibRr2FHMks5tjHJ5F8dMZh3AcwJEMf5vlfC0lxk=
-github.com/tidwall/gjson v1.18.0 h1:FIDeeyB800efLX89e5a8Y0BNH+LOngJyGrIWxG2FKQY=
-github.com/tidwall/gjson v1.18.0/go.mod h1:/wbyibRr2FHMks5tjHJ5F8dMZh3AcwJEMf5vlfC0lxk=
+github.com/tidwall/gjson v1.19.0 h1:xwxm7n691Uf3u5OFjzngavjGTh55KX5q/9w9xHW88JU=
+github.com/tidwall/gjson v1.19.0/go.mod h1:V37/opeE/JbLUOfH0QTXiNez2l0RUjYUhpT4szFQAfc=
 github.com/tidwall/match v1.1.1/go.mod h1:eRSPERbgtNPcGhD8UCthc6PmLEQXEWd3PRB5JTxsfmM=
 github.com/tidwall/match v1.2.0 h1:0pt8FlkOwjN2fPt4bIl4BoNxb98gGHN2ObFEDkrfZnM=
 github.com/tidwall/match v1.2.0/go.mod h1:eRSPERbgtNPcGhD8UCthc6PmLEQXEWd3PRB5JTxsfmM=
@@ -276,54 +278,54 @@ github.com/yuin/goldmark v1.8.2 h1:kEGpgqJXdgbkhcOgBxkC0X0PmoPG1ZyoZ117rDVp4zE=
 github.com/yuin/goldmark v1.8.2/go.mod h1:ip/1k0VRfGynBgxOz0yCqHrbZXhcjxyuS66Brc7iBKg=
 go.opentelemetry.io/auto/sdk v1.2.1 h1:jXsnJ4Lmnqd11kwkBV2LgLoFMZKizbCi5fNZ/ipaZ64=
 go.opentelemetry.io/auto/sdk v1.2.1/go.mod h1:KRTj+aOaElaLi+wW1kO/DZRXwkF4C5xPbEe3ZiIhN7Y=
-go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.68.0 h1:0Qx7VGBacMm9ZENQ7TnNObTYI4ShC+lHI16seduaxZo=
-go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.68.0/go.mod h1:Sje3i3MjSPKTSPvVWCaL8ugBzJwik3u4smCjUeuupqg=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0 h1:CqXxU8VOmDefoh0+ztfGaymYbhdB/tT3zs79QaZTNGY=
-go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.68.0/go.mod h1:BuhAPThV8PBHBvg8ZzZ/Ok3idOdhWIodywz2xEcRbJo=
-go.opentelemetry.io/otel v1.43.0 h1:mYIM03dnh5zfN7HautFE4ieIig9amkNANT+xcVxAj9I=
-go.opentelemetry.io/otel v1.43.0/go.mod h1:JuG+u74mvjvcm8vj8pI5XiHy1zDeoCS2LB1spIq7Ay0=
-go.opentelemetry.io/otel/metric v1.43.0 h1:d7638QeInOnuwOONPp4JAOGfbCEpYb+K6DVWvdxGzgM=
-go.opentelemetry.io/otel/metric v1.43.0/go.mod h1:RDnPtIxvqlgO8GRW18W6Z/4P462ldprJtfxHxyKd2PY=
-go.opentelemetry.io/otel/sdk v1.43.0 h1:pi5mE86i5rTeLXqoF/hhiBtUNcrAGHLKQdhg4h4V9Dg=
-go.opentelemetry.io/otel/sdk v1.43.0/go.mod h1:P+IkVU3iWukmiit/Yf9AWvpyRDlUeBaRg6Y+C58QHzg=
-go.opentelemetry.io/otel/sdk/metric v1.43.0 h1:S88dyqXjJkuBNLeMcVPRFXpRw2fuwdvfCGLEo89fDkw=
-go.opentelemetry.io/otel/sdk/metric v1.43.0/go.mod h1:C/RJtwSEJ5hzTiUz5pXF1kILHStzb9zFlIEe85bhj6A=
-go.opentelemetry.io/otel/trace v1.43.0 h1:BkNrHpup+4k4w+ZZ86CZoHHEkohws8AY+WTX09nk+3A=
-go.opentelemetry.io/otel/trace v1.43.0/go.mod h1:/QJhyVBUUswCphDVxq+8mld+AvhXZLhe+8WVFxiFff0=
+go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.69.0 h1:2yEATaop1/a1I4psnSLgWVPLWwCzkqWakgJy7xTDVy0=
+go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.69.0/go.mod h1:D7J12YRapIekYyPWgGPlA/23pRmpSEZC5xJC/TTLI9U=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.69.0 h1:8tvICD4vSTOOsNrsI4Ljf6C+6UKvpTEH5XY3JMoyPoo=
+go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.69.0/go.mod h1:z9+yiacE0IHRqM4qFfkbt/JYlmYXgss8GY/jXoNuPJI=
+go.opentelemetry.io/otel v1.44.0 h1:JjwHmHpA4iZ3wBxluu2fbbE7j4kqlE8jXyAyPXH7HqU=
+go.opentelemetry.io/otel v1.44.0/go.mod h1:BMgjTHL9WPRlRjL2oZCBTL4whCGtXch2H4BhOPIAyYc=
+go.opentelemetry.io/otel/metric v1.44.0 h1:1w0gILTcHdr3YI+ixLyjemwrVnsMURbTZFrSYCdDdmc=
+go.opentelemetry.io/otel/metric v1.44.0/go.mod h1:8O7hanEPBNgEMmybD3s2VBKcgWOCsA6tzHBPODAiquo=
+go.opentelemetry.io/otel/sdk v1.44.0 h1:nHYwb9lK+fJPU/dnT6s7W7Z8itMWyqrnVfbheVYrZ58=
+go.opentelemetry.io/otel/sdk v1.44.0/go.mod h1:Osuydd3Se74nqjAKxid74N5eC+jfEqfTegHRnq58oK0=
+go.opentelemetry.io/otel/sdk/metric v1.44.0 h1:3LlKgI+VjbVsjNRFZJZAJ30WjXC5VkNRks6si09iEfI=
+go.opentelemetry.io/otel/sdk/metric v1.44.0/go.mod h1:5B5pMARnXxKhltooO4xUuCBorl65a4EpnTalObqOigA=
+go.opentelemetry.io/otel/trace v1.44.0 h1:jxF5CsGYCe74MCRx2X4g7WsY/VBKRqqpNvXlX/6gtIk=
+go.opentelemetry.io/otel/trace v1.44.0/go.mod h1:oLl1jrMQAVo6v3GAggN+1VH9VIz9iUSvW53sW1Q8PIE=
 go.yaml.in/yaml/v3 v3.0.4 h1:tfq32ie2Jv2UxXFdLJdh3jXuOzWiL1fo0bu/FbuKpbc=
 go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
-golang.org/x/crypto v0.50.0 h1:zO47/JPrL6vsNkINmLoo/PH1gcxpls50DNogFvB5ZGI=
-golang.org/x/crypto v0.50.0/go.mod h1:3muZ7vA7PBCE6xgPX7nkzzjiUq87kRItoJQM1Yo8S+Q=
-golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f h1:W3F4c+6OLc6H2lb//N1q4WpJkhzJCK5J6kUi1NTVXfM=
-golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f/go.mod h1:J1xhfL/vlindoeF/aINzNzt2Bket5bjo9sdOYzOsU80=
-golang.org/x/net v0.53.0 h1:d+qAbo5L0orcWAr0a9JweQpjXF19LMXJE8Ey7hwOdUA=
-golang.org/x/net v0.53.0/go.mod h1:JvMuJH7rrdiCfbeHoo3fCQU24Lf5JJwT9W3sJFulfgs=
+golang.org/x/crypto v0.52.0 h1:RMs7fP2rXdep0CftQlK8Uf+kibLm7qkCcradZWYz988=
+golang.org/x/crypto v0.52.0/go.mod h1:1QgfPxDqh0T2M/elOJtp9RvuR95kVjir0e6/BvEmGbc=
+golang.org/x/exp v0.0.0-20260528193900-50dc527dd6c7 h1:cHpkPjp4TILjdZxz/O4ykwCpeS+dDqNuDGse4zgQDCk=
+golang.org/x/exp v0.0.0-20260528193900-50dc527dd6c7/go.mod h1:d2fgXJLVs4dYDHUk5lwMIfzRzSrWCfGZb0ZqeLa/Vcw=
+golang.org/x/net v0.55.0 h1:bcvxaJn3e1U6InsFWt1JUq1aSjnRxLzT2rtD2KfkDF8=
+golang.org/x/net v0.55.0/go.mod h1:L5U2KuzuOe1lY7Z+aWVIKK6qEeJXnXV9yzGA+WCHJww=
 golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs=
 golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q=
 golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
 golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
-golang.org/x/sys v0.43.0 h1:Rlag2XtaFTxp19wS8MXlJwTvoh8ArU6ezoyFsMyCTNI=
-golang.org/x/sys v0.43.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
-golang.org/x/term v0.42.0 h1:UiKe+zDFmJobeJ5ggPwOshJIVt6/Ft0rcfrXZDLWAWY=
-golang.org/x/term v0.42.0/go.mod h1:Dq/D+snpsbazcBG5+F9Q1n2rXV8Ma+71xEjTRufARgY=
-golang.org/x/text v0.36.0 h1:JfKh3XmcRPqZPKevfXVpI1wXPTqbkE5f7JA92a55Yxg=
-golang.org/x/text v0.36.0/go.mod h1:NIdBknypM8iqVmPiuco0Dh6P5Jcdk8lJL0CUebqK164=
+golang.org/x/sys v0.45.0 h1:dO4czNzziLiiXplLQgBCEpCvXQ3dnkn0SdaZSYdQ+FY=
+golang.org/x/sys v0.45.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/term v0.43.0 h1:S4RLU2sB31O/NCl+zFN9Aru9A/Cq2aqKpTZJ6B+DwT4=
+golang.org/x/term v0.43.0/go.mod h1:lrhlHNdQJHO+1qVYiHfFKVuVioJIheAc3fBSMFYEIsk=
+golang.org/x/text v0.37.0 h1:Cqjiwd9eSg8e0QAkyCaQTNHFIIzWtidPahFWR83rTrc=
+golang.org/x/text v0.37.0/go.mod h1:a5sjxXGs9hsn/AJVwuElvCAo9v8QYLzvavO5z2PiM38=
 golang.org/x/time v0.15.0 h1:bbrp8t3bGUeFOx08pvsMYRTCVSMk89u4tKbNOZbp88U=
 golang.org/x/time v0.15.0/go.mod h1:Y4YMaQmXwGQZoFaVFk4YpCt4FLQMYKZe9oeV/f4MSno=
 gonum.org/v1/gonum v0.17.0 h1:VbpOemQlsSMrYmn7T2OUvQ4dqxQXU+ouZFQsZOx50z4=
 gonum.org/v1/gonum v0.17.0/go.mod h1:El3tOrEuMpv2UdMrbNlKEh9vd86bmQ6vqIcDwxEOc1E=
-google.golang.org/api v0.277.0 h1:HJfyJUiNeBBUMai7ez8u14wkp/gH/I4wpGbbO9o+cSk=
-google.golang.org/api v0.277.0/go.mod h1:B9TqLBwJqVjp1mtt7WeoQwWRwvu/400y5lETOql+giQ=
-google.golang.org/genai v1.55.0 h1:iLHGk4Bj/IZ/GNNZb7hYqwSJMRBvqLeu2Hb6YQ+rYGw=
-google.golang.org/genai v1.55.0/go.mod h1:A3kkl0nyBjyFlNjgxIwKq70julKbIxpSxqKO5gw/gmk=
-google.golang.org/genproto v0.0.0-20260427160629-7cedc36a6bc4 h1:2iMJZntwvmfgtse+s744JY7v7PgEdSBuFYXucvpOHNM=
-google.golang.org/genproto v0.0.0-20260427160629-7cedc36a6bc4/go.mod h1:v14kaaboYyXQ1Gsu489Q+Hg/oN4B33mWtuOhF1HCeXA=
-google.golang.org/genproto/googleapis/api v0.0.0-20260427160629-7cedc36a6bc4 h1:yOzSCGPx+cp5VO7IxvZ9SBFF7j1tZVcNtlHR2iYKtVo=
-google.golang.org/genproto/googleapis/api v0.0.0-20260427160629-7cedc36a6bc4/go.mod h1:Q9HWtNeE7tM9npdIsEvqXj1QJIvVoeAV3rtXtS715Cw=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260427160629-7cedc36a6bc4 h1:tEkOQcXgF6dH1G+MVKZrfpYvozGrzb91k6ha7jireSM=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260427160629-7cedc36a6bc4/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
-google.golang.org/grpc v1.81.0 h1:W3G9N3KQf3BU+YuCtGKJk0CmxQNbAISICD/9AORxLIw=
-google.golang.org/grpc v1.81.0/go.mod h1:xGH9GfzOyMTGIOXBJmXt+BX/V0kcdQbdcuwQ/zNw42I=
+google.golang.org/api v0.282.0 h1:WmJiSVqUnKqJCpJOx7YADbXaC+9DDsnGSfllFSj7R2I=
+google.golang.org/api v0.282.0/go.mod h1:6Wssta4c5n9qHq5CBhmlai5h/PUa1djdDAIhYEHyvcM=
+google.golang.org/genai v1.58.0 h1:MNA3ZkRyr7MnRwZ9RNZ60p4+UMKV3yYRw6pyHq4pp0U=
+google.golang.org/genai v1.58.0/go.mod h1:A3kkl0nyBjyFlNjgxIwKq70julKbIxpSxqKO5gw/gmk=
+google.golang.org/genproto v0.0.0-20260504160031-60b97b32f348 h1:JjVGDZYWkJWZcxveJGzfkXC5myDVWAd4dZdgbzrDUv8=
+google.golang.org/genproto v0.0.0-20260504160031-60b97b32f348/go.mod h1:95PqD4xM+AdOcBGsmgfaofXsiA37uXDtDufVbntT3TU=
+google.golang.org/genproto/googleapis/api v0.0.0-20260504160031-60b97b32f348 h1:U8orV30l6KpDsi9dxU0CoJZGbjS8EEpw+6ba+XwGPQA=
+google.golang.org/genproto/googleapis/api v0.0.0-20260504160031-60b97b32f348/go.mod h1:Yzdzr5OOZFgSsEV2D/Xi9NL3bszpXFAg0hFJiRohcD8=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260526163538-3dc84a4a5aaa h1:mZHHdPZl0dbGHCflZgAq/Q468DWVFcU2whhB2KAo8fk=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260526163538-3dc84a4a5aaa/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
+google.golang.org/grpc v1.81.1 h1:VnnIIZ88UzOOKLukQi+ImGz8O1Wdp8nAGGnvOfEIWQQ=
+google.golang.org/grpc v1.81.1/go.mod h1:xGH9GfzOyMTGIOXBJmXt+BX/V0kcdQbdcuwQ/zNw42I=
 google.golang.org/protobuf v1.36.11 h1:fV6ZwhNocDyBLK0dj+fg8ektcVegBBuEolpbTQyBNVE=
 google.golang.org/protobuf v1.36.11/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

internal/acpserver/session.go

@@ -7,6 +7,7 @@ import (
 	"sync"
 
 	"github.com/charmbracelet/log"
+	"github.com/spf13/viper"
 
 	"github.com/mark3labs/kit/internal/extbridge"
 	"github.com/mark3labs/kit/internal/extensions"
@@ -38,10 +39,21 @@ func newSessionRegistry() *sessionRegistry {
 // given working directory. The Kit-generated session ID is used as the ACP
 // session ID so the mapping is 1:1.
 func (r *sessionRegistry) create(ctx context.Context, cwd string) (*acpSession, error) {
+	// Each ACP session gets its own isolated config store (CLI is left nil) so
+	// per-session SetModel / SetThinkingLevel calls cannot race or bleed across
+	// the sessionRegistry. We seed the relevant root-command flag values from
+	// the process-global store (which cobra populated from flags) so launching
+	// `kit acp -m <model> [--thinking-level ...] [--provider-url ...]` is still
+	// honored; .kit.yml and KIT_* env vars are loaded per session by kit.New.
+	streamOn := true
 	kitInstance, err := kit.New(ctx, &kit.Options{
-		SessionDir: cwd,
-		Quiet:      true,
-		Streaming:  true,
+		SessionDir:     cwd,
+		Quiet:          true,
+		Streaming:      &streamOn,
+		Model:          viper.GetString("model"),
+		ThinkingLevel:  viper.GetString("thinking-level"),
+		ProviderURL:    viper.GetString("provider-url"),
+		ProviderAPIKey: viper.GetString("provider-api-key"),
 	})
 	if err != nil {
 		// Provide actionable guidance for provider auth errors, which are

internal/agent/agent.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"io"
 	"strings"
+	"sync"
 	"time"
 
 	"charm.land/fantasy"
@@ -245,6 +246,12 @@ type Agent struct {
 	mcpReady chan struct{}
 	// mcpErr holds any error from background MCP loading.
 	mcpErr error
+
+	// promptMu serializes runtime updates to systemPrompt and the
+	// accompanying fantasy agent rebuild so concurrent SetSystemPrompt
+	// callers (e.g. Kit.applyComposedSystemPrompt invoked from multiple
+	// goroutines) don't race on a.systemPrompt / a.fantasyAgent.
+	promptMu sync.Mutex
 }
 
 // GenerateWithLoopResult contains the result and conversation history from an agent interaction.
@@ -585,8 +592,13 @@ func (a *Agent) GenerateWithCallbacks(ctx context.Context, messages []fantasy.Me
 	// This avoids type conflicts with provider-level options.
 	history = applyCacheControlToMessages(history)
 
-	// Track current tool call args for callbacks
-	var currentToolArgs string
+	// Track tool call args per-ToolCallID so parallel tool calls in a single
+	// step don't clobber each other. Without this, OnToolResult callbacks would
+	// all see the args of the last OnToolCall in the step. The mutex guards
+	// against the possibility that the underlying streaming layer dispatches
+	// callbacks from multiple goroutines.
+	toolCallArgs := make(map[string]string)
+	var toolCallArgsMu sync.Mutex
 
 	// Use the streaming path when streaming is enabled OR when any callbacks are
 	// provided. The agent only exposes tool/step callbacks on AgentStreamCall, so
@@ -773,7 +785,9 @@ func (a *Agent) GenerateWithCallbacks(ctx context.Context, messages []fantasy.Me
 				if ctx.Err() != nil {
 					return ctx.Err()
 				}
-				currentToolArgs = tc.Input
+				toolCallArgsMu.Lock()
+				toolCallArgs[tc.ToolCallID] = tc.Input
+				toolCallArgsMu.Unlock()
 
 				// Notify about the tool call
 				if cb.OnToolCall != nil {
@@ -793,15 +807,22 @@ func (a *Agent) GenerateWithCallbacks(ctx context.Context, messages []fantasy.Me
 				if ctx.Err() != nil {
 					return ctx.Err()
 				}
+				// Look up the args recorded for this specific tool call. Delete
+				// the entry so the map doesn't accumulate across steps.
+				toolCallArgsMu.Lock()
+				args := toolCallArgs[tr.ToolCallID]
+				delete(toolCallArgs, tr.ToolCallID)
+				toolCallArgsMu.Unlock()
+
 				// Notify tool execution finished
 				if cb.OnToolExecution != nil {
-					cb.OnToolExecution(tr.ToolCallID, tr.ToolName, currentToolArgs, false)
+					cb.OnToolExecution(tr.ToolCallID, tr.ToolName, args, false)
 				}
 
 				if cb.OnToolResult != nil {
 					// Extract result text and error status
 					resultText, isError := extractToolResultText(tr)
-					cb.OnToolResult(tr.ToolCallID, tr.ToolName, currentToolArgs, resultText, tr.ClientMetadata, isError)
+					cb.OnToolResult(tr.ToolCallID, tr.ToolName, args, resultText, tr.ClientMetadata, isError)
 				}
 
 				return nil
@@ -1303,6 +1324,24 @@ func (a *Agent) GetModel() fantasy.LanguageModel {
 	return a.model
 }
 
+// SetSystemPrompt updates the agent's system prompt and rebuilds the underlying
+// fantasy agent so subsequent turns use the new prompt. Safe to call while the
+// agent is idle; if invoked during an in-flight turn the new prompt takes
+// effect on the next LLM call.
+func (a *Agent) SetSystemPrompt(prompt string) {
+	a.promptMu.Lock()
+	defer a.promptMu.Unlock()
+	a.systemPrompt = prompt
+	a.rebuildFantasyAgent()
+}
+
+// GetSystemPrompt returns the agent's current system prompt.
+func (a *Agent) GetSystemPrompt() string {
+	a.promptMu.Lock()
+	defer a.promptMu.Unlock()
+	return a.systemPrompt
+}
+
 // GetMaxTokens returns the effective max output tokens the agent currently
 // sends to the LLM provider, after per-model defaults, right-sizing, and any
 // Anthropic thinking-budget adjustments. Returns 0 when no ModelConfig is

internal/agent/agent_parallel_tool_args_test.go

@@ -0,0 +1,109 @@
+package agent
+
+import (
+	"context"
+	"sync"
+	"testing"
+
+	"charm.land/fantasy"
+)
+
+// fakeParallelAgent simulates a provider that emits two parallel tool_use
+// blocks in a single step. It invokes the streaming callbacks in the order:
+//
+//	OnToolCall(A) -> OnToolCall(B) -> OnToolResult(A) -> OnToolResult(B)
+//
+// Before the fix in #33 the agent-layer wrapper recorded a single
+// `currentToolArgs` variable that was clobbered by the second OnToolCall, so
+// both OnToolResult callbacks received B's args instead of their own.
+type fakeParallelAgent struct {
+	calls   []fantasy.ToolCallContent
+	results []fantasy.ToolResultContent
+}
+
+func (f *fakeParallelAgent) Generate(_ context.Context, _ fantasy.AgentCall) (*fantasy.AgentResult, error) {
+	return &fantasy.AgentResult{}, nil
+}
+
+func (f *fakeParallelAgent) Stream(_ context.Context, opts fantasy.AgentStreamCall) (*fantasy.AgentResult, error) {
+	for _, tc := range f.calls {
+		if opts.OnToolCall != nil {
+			if err := opts.OnToolCall(tc); err != nil {
+				return nil, err
+			}
+		}
+	}
+	for _, tr := range f.results {
+		if opts.OnToolResult != nil {
+			if err := opts.OnToolResult(tr); err != nil {
+				return nil, err
+			}
+		}
+	}
+	return &fantasy.AgentResult{}, nil
+}
+
+// TestGenerateWithCallbacks_ParallelToolArgs is the regression test for #33.
+// It drives the streaming-callback wiring inside GenerateWithCallbacks with a
+// fake fantasy.Agent that emits two parallel tool calls before either result.
+// Each OnToolResult must receive the args of its own tool call (matched by
+// ToolCallID), not the args of the last OnToolCall in the step.
+func TestGenerateWithCallbacks_ParallelToolArgs(t *testing.T) {
+	t.Parallel()
+
+	argsA := `{"name":"scheduled_jobs"}`
+	argsB := `{"name":"gmail_trigger"}`
+
+	fake := &fakeParallelAgent{
+		calls: []fantasy.ToolCallContent{
+			{ToolCallID: "kit-A", ToolName: "load_skill", Input: argsA},
+			{ToolCallID: "kit-B", ToolName: "load_skill", Input: argsB},
+		},
+		results: []fantasy.ToolResultContent{
+			{ToolCallID: "kit-A", ToolName: "load_skill", Result: fantasy.ToolResultOutputContentText{Text: "ok-A"}},
+			{ToolCallID: "kit-B", ToolName: "load_skill", Result: fantasy.ToolResultOutputContentText{Text: "ok-B"}},
+		},
+	}
+
+	a := &Agent{
+		fantasyAgent:     fake,
+		streamingEnabled: false, // exercise the "hasCallbacks" branch
+	}
+
+	var mu sync.Mutex
+	resultArgs := map[string]string{}
+	executionArgs := map[string]string{} // captured when running == false
+
+	cb := GenerateCallbacks{
+		OnToolExecution: func(id, _, args string, running bool) {
+			if running {
+				return
+			}
+			mu.Lock()
+			defer mu.Unlock()
+			executionArgs[id] = args
+		},
+		OnToolResult: func(id, _, args, _, _ string, _ bool) {
+			mu.Lock()
+			defer mu.Unlock()
+			resultArgs[id] = args
+		},
+	}
+
+	if _, err := a.GenerateWithCallbacks(context.Background(), nil, cb); err != nil {
+		t.Fatalf("GenerateWithCallbacks returned error: %v", err)
+	}
+
+	if got, want := resultArgs["kit-A"], argsA; got != want {
+		t.Errorf("OnToolResult for kit-A: args = %q, want %q", got, want)
+	}
+	if got, want := resultArgs["kit-B"], argsB; got != want {
+		t.Errorf("OnToolResult for kit-B: args = %q, want %q", got, want)
+	}
+	if got, want := executionArgs["kit-A"], argsA; got != want {
+		t.Errorf("OnToolExecution(finish) for kit-A: args = %q, want %q", got, want)
+	}
+	if got, want := executionArgs["kit-B"], argsB; got != want {
+		t.Errorf("OnToolExecution(finish) for kit-B: args = %q, want %q", got, want)
+	}
+}

internal/agent/mcp_adapter.go

@@ -9,12 +9,19 @@ import (
 	"github.com/mark3labs/kit/internal/tools"
 )
 
+// mcpExecutor is the subset of *tools.MCPToolManager that the adapter
+// actually uses. Extracted as an interface so the adapter is unit-testable
+// without constructing a full manager + connection pool.
+type mcpExecutor interface {
+	ExecuteTool(ctx context.Context, prefixedName, inputJSON string) (*tools.MCPToolResult, error)
+}
+
 // mcpAgentTool adapts an tools.MCPTool to the fantasy.AgentTool interface.
 // This keeps the fantasy dependency confined to the agent layer — the tools
 // package is a pure MCP client library with no LLM framework dependency.
 type mcpAgentTool struct {
 	tool            tools.MCPTool
-	manager         *tools.MCPToolManager
+	exec            mcpExecutor
 	providerOptions fantasy.ProviderOptions
 }
 
@@ -29,10 +36,26 @@ func (t *mcpAgentTool) Info() fantasy.ToolInfo {
 }
 
 // Run executes the MCP tool by delegating to the MCPToolManager.
+//
+// MCP-side failures (JSON-RPC protocol errors, transport failures, schema
+// validation rejections from the server) are surfaced to the model as soft
+// tool errors rather than escalated to a critical agent error. This matches
+// the contract that native Kit tools follow via kit.ErrorResult(...) and
+// lets the model self-correct (e.g. retry with a fixed argument shape) or
+// give up gracefully rather than aborting the turn mid-run.
+//
+// Context cancellation is the one exception: if the caller cancelled the
+// context the turn was aborted intentionally, so we propagate the ctx error
+// to let the agent loop unwind cleanly.
 func (t *mcpAgentTool) Run(ctx context.Context, call fantasy.ToolCall) (fantasy.ToolResponse, error) {
-	result, err := t.manager.ExecuteTool(ctx, t.tool.Name, call.Input)
+	result, err := t.exec.ExecuteTool(ctx, t.tool.Name, call.Input)
 	if err != nil {
-		return fantasy.ToolResponse{}, fmt.Errorf("mcp tool execution failed: %w", err)
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return fantasy.ToolResponse{}, ctxErr
+		}
+		return fantasy.NewTextErrorResponse(
+			fmt.Sprintf("MCP tool %q failed: %s", t.tool.Name, err.Error()),
+		), nil
 	}
 
 	if result.IsError {
@@ -57,8 +80,8 @@ func mcpToolsToAgentTools(mcpTools []tools.MCPTool, manager *tools.MCPToolManage
 	agentTools := make([]fantasy.AgentTool, len(mcpTools))
 	for i, t := range mcpTools {
 		agentTools[i] = &mcpAgentTool{
-			tool:    t,
-			manager: manager,
+			tool: t,
+			exec: manager,
 		}
 	}
 	return agentTools

internal/agent/mcp_adapter_test.go

@@ -0,0 +1,158 @@
+package agent
+
+import (
+	"context"
+	"errors"
+	"strings"
+	"testing"
+	"time"
+
+	"charm.land/fantasy"
+
+	"github.com/mark3labs/kit/internal/tools"
+)
+
+// stubExecutor lets each test script the (result, err) pair returned by
+// ExecuteTool. The adapter holds an mcpExecutor interface, so this is the
+// only seam the tests need.
+type stubExecutor struct {
+	result *tools.MCPToolResult
+	err    error
+	// called records the last invocation for assertion.
+	called bool
+	name   string
+	input  string
+}
+
+func (s *stubExecutor) ExecuteTool(_ context.Context, prefixedName, inputJSON string) (*tools.MCPToolResult, error) {
+	s.called = true
+	s.name = prefixedName
+	s.input = inputJSON
+	return s.result, s.err
+}
+
+func newMCPAgentTool(exec mcpExecutor, name string) *mcpAgentTool {
+	return &mcpAgentTool{
+		tool: tools.MCPTool{Name: name},
+		exec: exec,
+	}
+}
+
+// Manager-side Go errors (JSON-RPC protocol errors, transport failures,
+// schema validation rejections from the MCP server) must be surfaced to
+// the model as soft tool errors so the agent loop can keep going. Aborting
+// the turn would discard all prior tool results — see issue #N.
+func TestMCPAgentTool_RPCErrorBecomesSoftError(t *testing.T) {
+	exec := &stubExecutor{
+		err: errors.New("MCP error -32602: Invalid params: missing field \"task\""),
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{
+		ID:    "call-1",
+		Name:  "pubmed__search",
+		Input: `{"query":"foo"}`,
+	})
+
+	if err != nil {
+		t.Fatalf("expected nil error (soft), got %v", err)
+	}
+	if !resp.IsError {
+		t.Fatalf("expected IsError=true, got false")
+	}
+	if !strings.Contains(resp.Content, "pubmed__search") {
+		t.Errorf("expected tool name in error content, got %q", resp.Content)
+	}
+	if !strings.Contains(resp.Content, "-32602") {
+		t.Errorf("expected underlying error text in content, got %q", resp.Content)
+	}
+}
+
+// Context cancellation is the one error that must remain critical: it
+// means the caller intentionally aborted, and the agent loop needs to
+// unwind cleanly rather than burning more steps.
+func TestMCPAgentTool_CtxCancelStaysCritical(t *testing.T) {
+	exec := &stubExecutor{
+		// Real managers typically return ctx.Err() (or a wrapper) when the
+		// context is cancelled mid-call.
+		err: context.Canceled,
+	}
+	tool := newMCPAgentTool(exec, "slow__tool")
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+
+	resp, err := tool.Run(ctx, fantasy.ToolCall{Name: "slow__tool"})
+
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context.Canceled, got %v", err)
+	}
+	if resp.IsError || resp.Content != "" {
+		t.Errorf("expected empty response on critical error, got IsError=%v Content=%q", resp.IsError, resp.Content)
+	}
+}
+
+// Deadline-exceeded behaves the same as cancellation: ctx.Err() is
+// non-nil, so the adapter must propagate the critical error rather than
+// converting the executor's error into a soft response.
+func TestMCPAgentTool_CtxDeadlineStaysCritical(t *testing.T) {
+	exec := &stubExecutor{err: context.DeadlineExceeded}
+	tool := newMCPAgentTool(exec, "slow__tool")
+
+	ctx, cancel := context.WithDeadline(context.Background(), time.Now().Add(-time.Second))
+	defer cancel()
+
+	resp, err := tool.Run(ctx, fantasy.ToolCall{Name: "slow__tool"})
+	if !errors.Is(err, context.DeadlineExceeded) {
+		t.Fatalf("expected context.DeadlineExceeded, got %v", err)
+	}
+	if resp.IsError || resp.Content != "" {
+		t.Errorf("expected empty response on critical error, got IsError=%v Content=%q", resp.IsError, resp.Content)
+	}
+}
+
+// Server-side soft errors (CallToolResult{ isError: true }) must continue
+// to flow through as soft errors — this was the existing behavior and
+// must not regress.
+func TestMCPAgentTool_ServerIsErrorRemainsSoftError(t *testing.T) {
+	exec := &stubExecutor{
+		result: &tools.MCPToolResult{
+			IsError: true,
+			Content: "search service is rate limited; try again in 30s",
+		},
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{Name: "pubmed__search"})
+	if err != nil {
+		t.Fatalf("expected nil error, got %v", err)
+	}
+	if !resp.IsError {
+		t.Fatalf("expected IsError=true, got false")
+	}
+	if resp.Content != "search service is rate limited; try again in 30s" {
+		t.Errorf("expected pass-through content, got %q", resp.Content)
+	}
+}
+
+// Happy path: ordinary successful tool result is passed through unchanged.
+func TestMCPAgentTool_SuccessIsPassthrough(t *testing.T) {
+	exec := &stubExecutor{
+		result: &tools.MCPToolResult{
+			IsError: false,
+			Content: `{"hits":3}`,
+		},
+	}
+	tool := newMCPAgentTool(exec, "pubmed__search")
+
+	resp, err := tool.Run(context.Background(), fantasy.ToolCall{Name: "pubmed__search"})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+	if resp.IsError {
+		t.Fatalf("expected IsError=false")
+	}
+	if resp.Content != `{"hits":3}` {
+		t.Errorf("expected pass-through content, got %q", resp.Content)
+	}
+}

internal/app/app.go

@@ -70,14 +70,17 @@ type App struct {
 	rootCtx    context.Context
 	rootCancel context.CancelFunc
 
-	// widgetUpdatePending is set to true when a WidgetUpdateEvent has been
-	// sent to the TUI but not yet consumed by its event loop. While the flag
-	// is set, subsequent NotifyWidgetUpdate calls are coalesced (dropped) to
-	// prevent fast extension tickers from flooding the BubbleTea mailbox with
-	// redundant re-render triggers. The flag is cleared after a short debounce
-	// (~1 frame) so new updates are always let through once the TUI has had a
-	// chance to process the pending event.
-	widgetUpdatePending atomic.Bool
+	// widgetUpdatePending is set to true while a WidgetUpdateEvent burst is
+	// being coalesced. The leading edge fires immediately; subsequent calls
+	// within the debounce window set widgetUpdateTrailing so a final event
+	// is delivered with the latest runner state at the end of the window.
+	// Without the trailing send, a rapid SetWidget→RemoveWidget pair (e.g.
+	// SubagentEnd pushing a final frame then removing the widget) would let
+	// the second call get silently dropped, leaving the TUI's layout stuck
+	// on the pre-removal widget height — visible as empty rows below the
+	// status bar after the widget disappears.
+	widgetUpdatePending  atomic.Bool
+	widgetUpdateTrailing atomic.Bool
 
 	// steerDrainFn is the test seam used by releaseBusyAfterCompact to pull
 	// any steer messages that arrived during compaction. In production it is
@@ -1157,32 +1160,47 @@ func (a *App) NotifyModelChanged(provider, model string) {
 // extension widgets. Called from the extension context's SetWidget/RemoveWidget
 // closures. In non-interactive mode this is a no-op (widgets are TUI-only).
 //
-// Coalescing: if a WidgetUpdateEvent is already queued and not yet consumed
-// by the TUI event loop, additional calls within the same ~16 ms window are
-// dropped. This prevents fast extension tickers from flooding BubbleTea's
-// mailbox with redundant re-render triggers.
+// Coalescing (leading + trailing edge): the first call in an idle period
+// fires immediately for responsiveness. Subsequent calls within a ~16 ms
+// debounce window are batched into a single trailing event delivered at
+// the end of the window. The trailing send is essential for correctness:
+// extensions routinely make tight SetWidget→RemoveWidget pairs (e.g. on
+// SubagentEnd) and silently dropping the second call would leave the TUI's
+// layout stuck on stale widget dimensions until some other event happens
+// to trigger a re-render.
 func (a *App) NotifyWidgetUpdate() {
-	// Coalesce: only one pending update at a time.
 	if !a.widgetUpdatePending.CompareAndSwap(false, true) {
+		// A leading-edge event is already in flight — mark that the runner
+		// state has changed again so the trailing send below picks it up.
+		a.widgetUpdateTrailing.Store(true)
 		return
 	}
 	a.mu.Lock()
 	prog := a.program
 	a.mu.Unlock()
-	if prog != nil {
-		prog.Send(WidgetUpdateEvent{})
-		// Reset the pending flag after a short debounce so subsequent calls
-		// within the same render cycle are also coalesced, but new updates
-		// after the cycle are allowed through.
-		go func() {
-			time.Sleep(16 * time.Millisecond) // ~1 frame at 60 fps
-			a.widgetUpdatePending.Store(false)
-		}()
-	} else {
+	if prog == nil {
 		// No program registered (non-interactive mode); clear the flag so
 		// future calls are never permanently blocked.
 		a.widgetUpdatePending.Store(false)
+		return
 	}
+	prog.Send(WidgetUpdateEvent{})
+	go func() {
+		time.Sleep(16 * time.Millisecond) // ~1 frame at 60 fps
+		// If any extra calls came in during the debounce window, deliver
+		// one trailing event so the TUI sees the latest widget state. We
+		// swap-and-test instead of plain-load so concurrent calls after
+		// the trailing send still race correctly with the pending reset.
+		if a.widgetUpdateTrailing.Swap(false) {
+			a.mu.Lock()
+			p := a.program
+			a.mu.Unlock()
+			if p != nil {
+				p.Send(WidgetUpdateEvent{})
+			}
+		}
+		a.widgetUpdatePending.Store(false)
+	}()
 }
 
 // NotifyContentReload sends a ContentReloadEvent to the TUI so it refreshes

internal/auth/credentials.go

@@ -255,29 +255,6 @@ func (cm *CredentialManager) HasAnthropicCredentials() (bool, error) {
 	}
 }
 
-// SetOpenAICredentials stores OpenAI API key credentials. It validates the
-// API key format before storing. The API key must start with "sk-" and be
-// at least 20 characters long. Returns an error if the API key is invalid or
-// if storage fails.
-func (cm *CredentialManager) SetOpenAICredentials(apiKey string) error {
-	if err := validateOpenAIAPIKey(apiKey); err != nil {
-		return err
-	}
-
-	store, err := cm.LoadCredentials()
-	if err != nil {
-		return err
-	}
-
-	store.OpenAI = &OpenAICredentials{
-		Type:      "api_key",
-		APIKey:    apiKey,
-		CreatedAt: time.Now(),
-	}
-
-	return cm.SaveCredentials(store)
-}
-
 // GetOpenAICredentials retrieves stored OpenAI credentials. Returns nil if
 // no credentials are stored. The returned credentials may be either OAuth or API
 // key type, check the Type field to determine which.
@@ -417,26 +394,6 @@ func validateAnthropicAPIKey(apiKey string) error {
 	return nil
 }
 
-// validateOpenAIAPIKey validates the format of an OpenAI API key
-func validateOpenAIAPIKey(apiKey string) error {
-	apiKey = strings.TrimSpace(apiKey)
-
-	if apiKey == "" {
-		return fmt.Errorf("API key cannot be empty")
-	}
-
-	// OpenAI API keys typically start with "sk-" and are quite long
-	if !strings.HasPrefix(apiKey, "sk-") {
-		return fmt.Errorf("invalid OpenAI API key format (should start with 'sk-')")
-	}
-
-	if len(apiKey) < 20 {
-		return fmt.Errorf("API key appears to be too short")
-	}
-
-	return nil
-}
-
 // GetAnthropicAPIKey retrieves an Anthropic API key from multiple sources in priority order:
 // 1. Command-line flag value (highest priority)
 // 2. Stored credentials (OAuth or API key)

internal/config/merger.go

@@ -7,32 +7,48 @@ import (
 	"github.com/spf13/viper"
 )
 
-// LoadAndValidateConfig loads configuration from viper, fixes environment variable
-// casing issues, and validates the configuration. Returns an error if loading or
-// validation fails.
+// LoadAndValidateConfig loads configuration from the process-global viper
+// store, fixes environment variable casing issues, and validates the
+// configuration. Returns an error if loading or validation fails.
+//
+// This is a convenience wrapper around [LoadAndValidateConfigFrom] using the
+// shared global store; it is retained for the CLI and other callers that rely
+// on viper's process-global state.
 func LoadAndValidateConfig() (*Config, error) {
+	return LoadAndValidateConfigFrom(viper.GetViper())
+}
+
+// LoadAndValidateConfigFrom loads configuration from the supplied per-instance
+// store, fixes environment variable casing issues, and validates the
+// configuration. When v is nil, the process-global store is used. Threading an
+// explicit store lets each Kit instance own an isolated configuration without
+// clobbering other instances in the same process.
+func LoadAndValidateConfigFrom(v *viper.Viper) (*Config, error) {
+	if v == nil {
+		v = viper.GetViper()
+	}
 	config := &Config{
 		MCPServers: make(map[string]MCPServerConfig),
 	}
-	if err := viper.Unmarshal(config); err != nil {
-		return nil, fmt.Errorf("failed to unmarshal config: %v", err)
+	if err := v.Unmarshal(config); err != nil {
+		return nil, fmt.Errorf("failed to unmarshal config: %w", err)
 	}
 
 	// Fix environment variable case sensitivity issue
 	// Viper lowercases all keys, but we need to preserve the original case for environment variables
-	fixEnvironmentCase(config)
+	fixEnvironmentCase(v, config)
 
 	if err := config.Validate(); err != nil {
-		return nil, fmt.Errorf("invalid config: %v", err)
+		return nil, fmt.Errorf("invalid config: %w", err)
 	}
 
 	return config, nil
 }
 
 // fixEnvironmentCase fixes the case of environment variable keys that were lowercased by Viper
-func fixEnvironmentCase(config *Config) {
+func fixEnvironmentCase(v *viper.Viper, config *Config) {
 	// Get the raw config data from viper
-	rawConfig := viper.AllSettings()
+	rawConfig := v.AllSettings()
 
 	// Check if we have mcpServers in the raw config
 	if mcpServersRaw, ok := rawConfig["mcpservers"]; ok {

internal/core/bash.go

@@ -160,15 +160,6 @@ func rewriteSudoForStdin(command string) string {
 	return result
 }
 
-// SudoPasswordRequiredResult is a special marker that indicates sudo needs a password.
-// This is stored in tool response metadata to signal the TUI to prompt for password.
-const SudoPasswordRequiredMetadata = `{"sudo_password_required":true}`
-
-// IsSudoPasswordRequiredResult checks if a tool response indicates sudo password is needed.
-func IsSudoPasswordRequiredResult(resp fantasy.ToolResponse) bool {
-	return resp.Metadata == SudoPasswordRequiredMetadata
-}
-
 func executeBash(ctx context.Context, call fantasy.ToolCall, workDir string) (fantasy.ToolResponse, error) {
 	var args bashArgs
 	if err := parseArgs(call.Input, &args); err != nil {

internal/extensions/runner.go

@@ -98,9 +98,20 @@ type Runner struct {
 	disabledTools   map[string]bool           // nil = all tools enabled
 	customEventSubs map[string][]func(string) // inter-extension event bus
 	optionOverrides map[string]string         // runtime option overrides
+	configStore     *viper.Viper              // per-instance config store (nil = global)
 	mu              sync.RWMutex
 }
 
+// SetConfigStore sets the per-instance configuration store used by GetOption
+// to resolve "options.<name>" config values. When unset (nil), GetOption falls
+// back to the process-global viper store. Threading a per-Kit store keeps
+// extension option resolution isolated between Kit instances.
+func (r *Runner) SetConfigStore(v *viper.Viper) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.configStore = v
+}
+
 // ShortcutEntry pairs a shortcut definition with its handler.
 type ShortcutEntry struct {
 	Def     ShortcutDef
@@ -872,7 +883,13 @@ func (r *Runner) GetOption(name string) string {
 
 	// 3. Viper config: options.<name>
 	configKey := "options." + name
-	if v := viper.GetString(configKey); v != "" {
+	r.mu.RLock()
+	store := r.configStore
+	r.mu.RUnlock()
+	if store == nil {
+		store = viper.GetViper()
+	}
+	if v := store.GetString(configKey); v != "" {
 		return v
 	}
 

internal/kitsetup/setup.go

@@ -46,9 +46,9 @@ type AgentSetupOptions struct {
 	ToolWrapper func([]fantasy.AgentTool) []fantasy.AgentTool
 
 	// ProviderConfig, when non-nil, is used directly instead of calling
-	// BuildProviderConfig(). Callers that already hold viperInitMu can
-	// pre-build this and release the lock before calling SetupAgent, so the
-	// slow agent/MCP initialisation runs concurrently with other New() calls.
+	// BuildProviderConfig(). Callers (e.g. Kit.New) pre-build this from their
+	// per-instance config store and pass it here, so the slow agent/MCP
+	// initialisation can run without further config reads.
 	ProviderConfig *models.ProviderConfig
 	// Debug enables debug logging. When zero-value, viper is consulted.
 	// Only meaningful when ProviderConfig is also set.
@@ -75,6 +75,11 @@ type AgentSetupOptions struct {
 	// MCPTaskConfig configures task-augmented tools/call execution. The
 	// zero value preserves historical synchronous-only behaviour.
 	MCPTaskConfig tools.MCPTaskConfig
+	// Viper is the per-instance configuration store. When set, it is used for
+	// any fallback config reads (debug, no-extensions, max-steps, stream,
+	// extension paths) and is attached to the extension runner. When nil, the
+	// process-global viper store is used.
+	Viper *viper.Viper
 }
 
 // AgentSetupResult bundles the created agent and any debug logger so the caller
@@ -87,57 +92,62 @@ type AgentSetupResult struct {
 	ExtRunner *extensions.Runner
 }
 
-// BuildProviderConfig creates a *models.ProviderConfig from the current viper
-// state. All entry points (root, script, SDK) converge through this function.
+// BuildProviderConfig creates a *models.ProviderConfig from the supplied viper
+// store (or the process-global store when v is nil). All entry points (root,
+// script, SDK) converge through this function.
 //
 // Generation parameter pointers (Temperature, TopP, etc.) are only set when
 // the user has explicitly configured them via CLI flag, environment variable,
 // or global config file. This allows per-model defaults from modelSettings
 // and customModels to fill in unset parameters downstream.
-func BuildProviderConfig() (*models.ProviderConfig, string, error) {
-	systemPrompt, err := config.LoadSystemPrompt(viper.GetString("system-prompt"))
+func BuildProviderConfig(v *viper.Viper) (*models.ProviderConfig, string, error) {
+	if v == nil {
+		v = viper.GetViper()
+	}
+	systemPrompt, err := config.LoadSystemPrompt(v.GetString("system-prompt"))
 	if err != nil {
 		return nil, "", fmt.Errorf("failed to load system prompt: %w", err)
 	}
 
-	numGPU := int32(viper.GetInt("num-gpu-layers"))
-	mainGPU := int32(viper.GetInt("main-gpu"))
+	numGPU := int32(v.GetInt("num-gpu-layers"))
+	mainGPU := int32(v.GetInt("main-gpu"))
 
 	cfg := &models.ProviderConfig{
-		ModelString:    viper.GetString("model"),
+		ModelString:    v.GetString("model"),
 		SystemPrompt:   systemPrompt,
-		ProviderAPIKey: viper.GetString("provider-api-key"),
-		ProviderURL:    viper.GetString("provider-url"),
-		MaxTokens:      viper.GetInt("max-tokens"),
-		StopSequences:  viper.GetStringSlice("stop-sequences"),
+		ProviderAPIKey: v.GetString("provider-api-key"),
+		ProviderURL:    v.GetString("provider-url"),
+		MaxTokens:      v.GetInt("max-tokens"),
+		StopSequences:  v.GetStringSlice("stop-sequences"),
 		NumGPU:         &numGPU,
 		MainGPU:        &mainGPU,
-		TLSSkipVerify:  viper.GetBool("tls-skip-verify"),
-		ThinkingLevel:  models.ParseThinkingLevel(viper.GetString("thinking-level")),
+		TLSSkipVerify:  v.GetBool("tls-skip-verify"),
+		ThinkingLevel:  models.ParseThinkingLevel(v.GetString("thinking-level")),
+		ConfigStore:    v,
 	}
 
 	// Only set generation parameter pointers when the user has explicitly
 	// provided a value. This leaves nil pointers for unset params, allowing
 	// per-model defaults (modelSettings / customModels params) to apply.
-	if viper.IsSet("temperature") {
-		v := float32(viper.GetFloat64("temperature"))
-		cfg.Temperature = &v
+	if v.IsSet("temperature") {
+		val := float32(v.GetFloat64("temperature"))
+		cfg.Temperature = &val
 	}
-	if viper.IsSet("top-p") {
-		v := float32(viper.GetFloat64("top-p"))
-		cfg.TopP = &v
+	if v.IsSet("top-p") {
+		val := float32(v.GetFloat64("top-p"))
+		cfg.TopP = &val
 	}
-	if viper.IsSet("top-k") {
-		v := int32(viper.GetInt("top-k"))
-		cfg.TopK = &v
+	if v.IsSet("top-k") {
+		val := int32(v.GetInt("top-k"))
+		cfg.TopK = &val
 	}
-	if viper.IsSet("frequency-penalty") {
-		v := float32(viper.GetFloat64("frequency-penalty"))
-		cfg.FrequencyPenalty = &v
+	if v.IsSet("frequency-penalty") {
+		val := float32(v.GetFloat64("frequency-penalty"))
+		cfg.FrequencyPenalty = &val
 	}
-	if viper.IsSet("presence-penalty") {
-		v := float32(viper.GetFloat64("presence-penalty"))
-		cfg.PresencePenalty = &v
+	if v.IsSet("presence-penalty") {
+		val := float32(v.GetFloat64("presence-penalty"))
+		cfg.PresencePenalty = &val
 	}
 
 	return cfg, systemPrompt, nil
@@ -149,14 +159,21 @@ func SetupAgent(ctx context.Context, opts AgentSetupOptions) (*AgentSetupResult,
 	var modelConfig *models.ProviderConfig
 	var systemPrompt string
 
+	// Resolve the config store: prefer the per-instance store, falling back to
+	// the process-global store.
+	v := opts.Viper
+	if v == nil {
+		v = viper.GetViper()
+	}
+
 	if opts.ProviderConfig != nil {
-		// Pre-built config supplied by caller (e.g. Kit.New after releasing
-		// viperInitMu). Use it directly — no viper reads needed here.
+		// Pre-built config supplied by caller (e.g. Kit.New after building the
+		// per-instance store). Use it directly — no viper reads needed here.
 		modelConfig = opts.ProviderConfig
 		systemPrompt = modelConfig.SystemPrompt
 	} else {
 		var err error
-		modelConfig, systemPrompt, err = BuildProviderConfig()
+		modelConfig, systemPrompt, err = BuildProviderConfig(v)
 		if err != nil {
 			return nil, err
 		}
@@ -164,13 +181,13 @@ func SetupAgent(ctx context.Context, opts AgentSetupOptions) (*AgentSetupResult,
 
 	// Resolve debug / no-extensions / max-steps / streaming: prefer explicit
 	// fields (set when ProviderConfig was pre-built) over viper fallback.
-	debugEnabled := opts.Debug || viper.GetBool("debug")
-	noExtensions := opts.NoExtensions || viper.GetBool("no-extensions")
+	debugEnabled := opts.Debug || v.GetBool("debug")
+	noExtensions := opts.NoExtensions || v.GetBool("no-extensions")
 	maxSteps := opts.MaxSteps
 	if maxSteps == 0 {
-		maxSteps = viper.GetInt("max-steps")
+		maxSteps = v.GetInt("max-steps")
 	}
-	streamingEnabled := opts.StreamingEnabled || viper.GetBool("stream")
+	streamingEnabled := opts.StreamingEnabled || v.GetBool("stream")
 
 	// Create the appropriate debug logger.
 	var debugLogger tools.DebugLogger
@@ -189,7 +206,7 @@ func SetupAgent(ctx context.Context, opts AgentSetupOptions) (*AgentSetupResult,
 	var extCreationOpts extensionCreationOpts
 	if !noExtensions {
 		var extErr error
-		extRunner, extCreationOpts, extErr = loadExtensions()
+		extRunner, extCreationOpts, extErr = loadExtensions(v)
 		if extErr != nil {
 			fmt.Printf("Warning: Failed to load extensions: %v\n", extErr)
 		}
@@ -253,9 +270,14 @@ type extensionCreationOpts struct {
 }
 
 // loadExtensions discovers and loads Yaegi extensions, builds the runner,
-// and returns the tool wrapper/extra tools.
-func loadExtensions() (*extensions.Runner, extensionCreationOpts, error) {
-	extraPaths := viper.GetStringSlice("extension")
+// and returns the tool wrapper/extra tools. The supplied store is used to
+// resolve the "extension" config key and is attached to the runner so
+// extension option lookups stay isolated to this Kit instance.
+func loadExtensions(v *viper.Viper) (*extensions.Runner, extensionCreationOpts, error) {
+	if v == nil {
+		v = viper.GetViper()
+	}
+	extraPaths := v.GetStringSlice("extension")
 	loaded, err := extensions.LoadExtensions(extraPaths)
 	if err != nil {
 		return nil, extensionCreationOpts{}, err
@@ -266,6 +288,7 @@ func loadExtensions() (*extensions.Runner, extensionCreationOpts, error) {
 	}
 
 	runner := extensions.NewRunner(loaded)
+	runner.SetConfigStore(v)
 
 	wrapper := func(tools []fantasy.AgentTool) []fantasy.AgentTool {
 		return extensions.WrapToolsWithExtensions(tools, runner)

internal/models/autoroute_test.go

@@ -0,0 +1,266 @@
+package models
+
+import (
+	"context"
+	"io"
+	"net/http"
+	"reflect"
+	"strings"
+	"testing"
+)
+
+// TestNpmToWireProtocol documents the wire protocols that the auto-router
+// understands. Provider-specific bundles (azure, bedrock, vercel, openrouter,
+// google-vertex*) are intentionally absent — they have native top-level cases
+// in CreateProvider and never reach the auto-router.
+func TestNpmToWireProtocol(t *testing.T) {
+	want := map[string]wireProtocol{
+		"@ai-sdk/openai":            wireOpenAI,
+		"@ai-sdk/openai-compatible": wireOpenAI,
+		"@ai-sdk/anthropic":         wireAnthropic,
+		"@ai-sdk/google":            wireGoogle,
+	}
+	for npm, wire := range want {
+		if got := npmToWireProtocol[npm]; got != wire {
+			t.Errorf("npmToWireProtocol[%q] = %d, want %d", npm, got, wire)
+		}
+	}
+
+	// Bundle packages must NOT be in the table (regression guard against the
+	// old npmToLLMProvider map that listed 10 entries but only handled 3).
+	for _, npm := range []string{
+		"@ai-sdk/google-vertex",
+		"@ai-sdk/google-vertex/anthropic",
+		"@ai-sdk/amazon-bedrock",
+		"@ai-sdk/azure",
+		"@openrouter/ai-sdk-provider",
+		"@ai-sdk/vercel",
+	} {
+		if _, ok := npmToWireProtocol[npm]; ok {
+			t.Errorf("npmToWireProtocol unexpectedly contains bundle package %q", npm)
+		}
+	}
+}
+
+// newTestRegistry builds a registry containing a single proxy-style provider
+// ("testproxy") with the given default npm, plus one model that carries the
+// given per-model npm override.
+func newTestRegistry(api, defaultNPM, modelID, modelNPMOverride string) *ModelsRegistry {
+	return &ModelsRegistry{
+		providers: map[string]ProviderInfo{
+			"testproxy": {
+				ID:   "testproxy",
+				Name: "Test Proxy",
+				Env:  []string{"TESTPROXY_API_KEY"},
+				NPM:  defaultNPM,
+				API:  api,
+				Models: map[string]ModelInfo{
+					modelID: {
+						ID:          modelID,
+						Name:        modelID,
+						ProviderNPM: modelNPMOverride,
+					},
+				},
+			},
+		},
+	}
+}
+
+// TestAutoRouteProvider_WireRouting verifies that autoRouteProvider routes each
+// npm package to the correct fantasy provider implementation. This is the core
+// regression test for issue #41: previously any npm that resolved to a
+// non-openai/anthropic/openaicompat LLM provider (notably @ai-sdk/google) hit a
+// dead `default` branch and failed with "has no LLM provider mapping".
+func TestAutoRouteProvider_WireRouting(t *testing.T) {
+	tests := []struct {
+		name        string
+		modelID     string
+		defaultNPM  string
+		overrideNPM string
+		// wantType is the concrete fantasy LanguageModel type the model should
+		// be routed to, identified by reflect type string.
+		wantType string
+	}{
+		{
+			name:       "openai-compatible default",
+			modelID:    "test-model",
+			defaultNPM: "@ai-sdk/openai-compatible",
+			wantType:   "openai.languageModel",
+		},
+		{
+			name:        "anthropic override",
+			modelID:     "test-model",
+			defaultNPM:  "@ai-sdk/openai-compatible",
+			overrideNPM: "@ai-sdk/anthropic",
+			wantType:    "anthropic.languageModel",
+		},
+		{
+			name:        "openai (responses) override",
+			modelID:     "gpt-4o",
+			defaultNPM:  "@ai-sdk/openai-compatible",
+			overrideNPM: "@ai-sdk/openai",
+			wantType:    "openai.responsesLanguageModel",
+		},
+		{
+			// The bug: opencode's gemini-* models override the default
+			// openai-compatible npm with @ai-sdk/google.
+			name:        "google override (issue #41)",
+			modelID:     "gemini-3.5-flash",
+			defaultNPM:  "@ai-sdk/openai-compatible",
+			overrideNPM: "@ai-sdk/google",
+			wantType:    "*google.languageModel",
+		},
+		{
+			// Unknown npm but provider has an API URL → openai-compatible fallback.
+			name:       "unknown npm with API URL falls back to openai-compat",
+			modelID:    "test-model",
+			defaultNPM: "@ai-sdk/some-future-thing",
+			wantType:   "openai.languageModel",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			reg := newTestRegistry("https://proxy.example/v1", tt.defaultNPM, tt.modelID, tt.overrideNPM)
+			config := &ProviderConfig{ProviderAPIKey: "test-key"}
+
+			result, err := autoRouteProvider(context.Background(), config, "testproxy", tt.modelID, reg)
+			if err != nil {
+				t.Fatalf("autoRouteProvider returned error: %v", err)
+			}
+			if result == nil || result.Model == nil {
+				t.Fatalf("autoRouteProvider returned nil model")
+			}
+
+			gotType := reflect.TypeOf(result.Model).String()
+			if gotType != tt.wantType {
+				t.Errorf("routed to %s, want %s", gotType, tt.wantType)
+			}
+		})
+	}
+}
+
+// TestAutoRouteProvider_UnknownNpmNoAPI verifies the improved error message for
+// a provider whose npm has no known wire protocol and that has no API URL to
+// fall back on.
+func TestAutoRouteProvider_UnknownNpmNoAPI(t *testing.T) {
+	reg := newTestRegistry("", "@ai-sdk/unmapped", "test-model", "")
+	config := &ProviderConfig{ProviderAPIKey: "test-key"}
+
+	_, err := autoRouteProvider(context.Background(), config, "testproxy", "test-model", reg)
+	if err == nil {
+		t.Fatal("expected error for unknown npm with no API URL, got nil")
+	}
+	if !strings.Contains(err.Error(), "cannot auto-route provider testproxy") {
+		t.Errorf("unexpected error message: %v", err)
+	}
+	if !strings.Contains(err.Error(), "--provider-url") {
+		t.Errorf("error should suggest --provider-url, got: %v", err)
+	}
+}
+
+// TestAutoRouteProvider_UnknownProvider verifies the not-in-database error.
+func TestAutoRouteProvider_UnknownProvider(t *testing.T) {
+	reg := newTestRegistry("https://proxy.example/v1", "@ai-sdk/openai-compatible", "test-model", "")
+	config := &ProviderConfig{ProviderAPIKey: "test-key"}
+
+	_, err := autoRouteProvider(context.Background(), config, "does-not-exist", "test-model", reg)
+	if err == nil {
+		t.Fatal("expected error for unknown provider, got nil")
+	}
+	if !strings.Contains(err.Error(), "not found in model database") {
+		t.Errorf("unexpected error message: %v", err)
+	}
+}
+
+// TestIsProviderLLMSupported_Google verifies that a provider whose npm is
+// @ai-sdk/google is reported as supported (it now maps to a wire protocol).
+func TestIsProviderLLMSupported_Google(t *testing.T) {
+	info := &ProviderInfo{ID: "testproxy", NPM: "@ai-sdk/google"}
+	if !isProviderLLMSupported("testproxy", info) {
+		t.Error("expected @ai-sdk/google provider to be LLM-supported")
+	}
+}
+
+// TestVersionedBasePath verifies detection of proxy base URLs that already
+// carry an API version segment (which collides with the genai SDK's injected
+// version).
+func TestVersionedBasePath(t *testing.T) {
+	tests := []struct {
+		rawURL string
+		want   string
+	}{
+		{"https://opencode.ai/zen/v1", "/zen/v1"},
+		{"https://opencode.ai/zen/v1/", "/zen/v1"},
+		{"https://example.com/api/v1beta", "/api/v1beta"},
+		{"https://example.com/api/v2alpha", "/api/v2alpha"},
+		{"https://generativelanguage.googleapis.com", ""},
+		{"https://proxy.example/openai", ""},
+		{"", ""},
+	}
+	for _, tt := range tests {
+		if got := versionedBasePath(tt.rawURL); got != tt.want {
+			t.Errorf("versionedBasePath(%q) = %q, want %q", tt.rawURL, got, tt.want)
+		}
+	}
+}
+
+// recordingRoundTripper captures the path of the request it receives.
+type recordingRoundTripper struct{ gotPath string }
+
+func (r *recordingRoundTripper) RoundTrip(req *http.Request) (*http.Response, error) {
+	r.gotPath = req.URL.Path
+	return &http.Response{
+		StatusCode: 200,
+		Body:       io.NopCloser(strings.NewReader("{}")),
+		Header:     make(http.Header),
+	}, nil
+}
+
+// TestGeminiProxyTransport_StripsInjectedVersion verifies that the transport
+// collapses the genai-injected "/v1beta" segment that follows a proxy base
+// URL which already carries its own version segment. This is the second-order
+// fix that makes opencode/gemini-* actually reach the proxy (issue #41).
+func TestGeminiProxyTransport_StripsInjectedVersion(t *testing.T) {
+	tests := []struct {
+		name     string
+		basePath string
+		reqPath  string
+		wantPath string
+	}{
+		{
+			name:     "strips doubled v1beta after /zen/v1",
+			basePath: "/zen/v1",
+			reqPath:  "/zen/v1/v1beta/models/gemini-3.5-flash:generateContent",
+			wantPath: "/zen/v1/models/gemini-3.5-flash:generateContent",
+		},
+		{
+			name:     "strips doubled v1beta1 after /zen/v1",
+			basePath: "/zen/v1",
+			reqPath:  "/zen/v1/v1beta1/models/gemini-3.5-flash:generateContent",
+			wantPath: "/zen/v1/models/gemini-3.5-flash:generateContent",
+		},
+		{
+			name:     "leaves non-matching path untouched",
+			basePath: "/zen/v1",
+			reqPath:  "/other/v1beta/models/x:generateContent",
+			wantPath: "/other/v1beta/models/x:generateContent",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			rec := &recordingRoundTripper{}
+			tr := &geminiProxyTransport{base: rec, basePath: tt.basePath}
+			req, err := http.NewRequest(http.MethodPost, "https://host"+tt.reqPath, nil)
+			if err != nil {
+				t.Fatalf("NewRequest: %v", err)
+			}
+			if _, err := tr.RoundTrip(req); err != nil {
+				t.Fatalf("RoundTrip: %v", err)
+			}
+			if rec.gotPath != tt.wantPath {
+				t.Errorf("forwarded path = %q, want %q", rec.gotPath, tt.wantPath)
+			}
+		})
+	}
+}

internal/models/custom.go

@@ -10,14 +10,24 @@ import (
 
 // loadCustomModelsFromConfig loads custom model definitions from the config file
 // and returns them as a map of model ID -> ModelInfo. Returns nil if no custom
-// models are configured.
+// models are configured. Reads from the process-global viper store (the model
+// registry is a process-global singleton).
 func loadCustomModelsFromConfig() map[string]ModelInfo {
-	if !viper.IsSet("customModels") {
+	return loadCustomModelsFrom(viper.GetViper())
+}
+
+// loadCustomModelsFrom loads custom model definitions from the supplied store.
+// When v is nil the process-global store is used.
+func loadCustomModelsFrom(v *viper.Viper) map[string]ModelInfo {
+	if v == nil {
+		v = viper.GetViper()
+	}
+	if !v.IsSet("customModels") {
 		return nil
 	}
 
 	var customModels map[string]CustomModelConfig
-	if err := viper.UnmarshalKey("customModels", &customModels); err != nil {
+	if err := v.UnmarshalKey("customModels", &customModels); err != nil {
 		log.Printf("Warning: Failed to parse customModels: %v", err)
 		return nil
 	}
@@ -60,15 +70,26 @@ func modelConfigToModelInfo(modelID string, cfg CustomModelConfig) ModelInfo {
 }
 
 // LoadModelSettingsFromConfig loads per-model generation parameter overrides
-// from the config file. Keys are "provider/model" strings. Returns nil if
-// no model settings are configured.
+// from the process-global viper store. Keys are "provider/model" strings.
+// Returns nil if no model settings are configured.
 func LoadModelSettingsFromConfig() map[string]*GenerationParams {
-	if !viper.IsSet("modelSettings") {
+	return LoadModelSettingsFrom(viper.GetViper())
+}
+
+// LoadModelSettingsFrom loads per-model generation parameter overrides from the
+// supplied per-instance store. When v is nil the process-global store is used.
+// Keys are "provider/model" strings. Returns nil if no model settings are
+// configured.
+func LoadModelSettingsFrom(v *viper.Viper) map[string]*GenerationParams {
+	if v == nil {
+		v = viper.GetViper()
+	}
+	if !v.IsSet("modelSettings") {
 		return nil
 	}
 
 	var settings map[string]GenerationParamsConfig
-	if err := viper.UnmarshalKey("modelSettings", &settings); err != nil {
+	if err := v.UnmarshalKey("modelSettings", &settings); err != nil {
 		log.Printf("Warning: Failed to parse modelSettings: %v", err)
 		return nil
 	}
@@ -148,12 +169,17 @@ func ApplyModelSettings(config *ProviderConfig, modelInfo *ModelInfo) {
 		return
 	}
 
+	// Resolve the config store: prefer the per-instance store carried on the
+	// ProviderConfig (set by BuildProviderConfig / Kit.New), falling back to
+	// the process-global store for callers that don't thread one through.
+	store := config.ConfigStore
+
 	// Collect model-level params: modelSettings override > custom model params.
 	// modelSettings takes priority because it's the more specific/intentional config.
 	var params *GenerationParams
 
 	// First check modelSettings from config.
-	if settings := LoadModelSettingsFromConfig(); settings != nil {
+	if settings := LoadModelSettingsFrom(store); settings != nil {
 		modelKey := provider + "/" + modelName
 		if p, ok := settings[modelKey]; ok {
 			params = p
@@ -173,28 +199,28 @@ func ApplyModelSettings(config *ProviderConfig, modelInfo *ModelInfo) {
 	// We check viper.IsSet() which returns true only when the key was
 	// set via CLI flag, environment variable, or config file global section.
 
-	if params.MaxTokens != nil && !isExplicitlySet("max-tokens") {
+	if params.MaxTokens != nil && !isExplicitlySet(store, "max-tokens") {
 		config.MaxTokens = *params.MaxTokens
 	}
-	if params.Temperature != nil && !isExplicitlySet("temperature") {
+	if params.Temperature != nil && !isExplicitlySet(store, "temperature") {
 		config.Temperature = params.Temperature
 	}
-	if params.TopP != nil && !isExplicitlySet("top-p") {
+	if params.TopP != nil && !isExplicitlySet(store, "top-p") {
 		config.TopP = params.TopP
 	}
-	if params.TopK != nil && !isExplicitlySet("top-k") {
+	if params.TopK != nil && !isExplicitlySet(store, "top-k") {
 		config.TopK = params.TopK
 	}
-	if params.FrequencyPenalty != nil && !isExplicitlySet("frequency-penalty") {
+	if params.FrequencyPenalty != nil && !isExplicitlySet(store, "frequency-penalty") {
 		config.FrequencyPenalty = params.FrequencyPenalty
 	}
-	if params.PresencePenalty != nil && !isExplicitlySet("presence-penalty") {
+	if params.PresencePenalty != nil && !isExplicitlySet(store, "presence-penalty") {
 		config.PresencePenalty = params.PresencePenalty
 	}
-	if len(params.StopSequences) > 0 && !isExplicitlySet("stop-sequences") {
+	if len(params.StopSequences) > 0 && !isExplicitlySet(store, "stop-sequences") {
 		config.StopSequences = params.StopSequences
 	}
-	if params.ThinkingLevel != "" && !isExplicitlySet("thinking-level") {
+	if params.ThinkingLevel != "" && !isExplicitlySet(store, "thinking-level") {
 		config.ThinkingLevel = params.ThinkingLevel
 	}
 	if params.SystemPrompt != "" && config.SystemPrompt == "" {
@@ -228,7 +254,14 @@ func LoadSystemPromptValue(input string) string {
 // isExplicitlySet returns true when the user has explicitly set a config key
 // via CLI flag, environment variable, or the global section of the config file.
 // Model-level defaults should not override explicitly set values.
-func isExplicitlySet(key string) bool {
+//
+// The check runs against the supplied per-instance store when non-nil,
+// otherwise the process-global store. This keeps the "explicit vs unset"
+// precedence contract per-Kit-instance once a store is threaded through.
+func isExplicitlySet(v *viper.Viper, key string) bool {
+	if v == nil {
+		v = viper.GetViper()
+	}
 	// viper.IsSet returns true if the key has been set in any of the
 	// data stores (flag, env, config file, default). We need to check
 	// whether the value was set at the global config level (not just
@@ -239,7 +272,7 @@ func isExplicitlySet(key string) bool {
 	// file values. This means global config file values (e.g.
 	// temperature: 0.7 at the top level) will correctly take precedence
 	// over model-level defaults, which is the desired behavior.
-	return viper.IsSet(key)
+	return v.IsSet(key)
 }
 
 // GenerationParams holds per-model generation parameter defaults.

internal/models/modelsdb.go

@@ -48,18 +48,28 @@ type modelsDBLimit struct {
 	Output  int `json:"output"`
 }
 
-// npmToLLMProvider maps npm package names from models.dev to LLM
-// provider identifiers. Providers not in this map but with an api URL
-// can be auto-routed through openaicompat.
-var npmToLLMProvider = map[string]string{
-	"@ai-sdk/anthropic":               "anthropic",
-	"@ai-sdk/openai":                  "openai",
-	"@ai-sdk/google":                  "google",
-	"@ai-sdk/google-vertex":           "google-vertex",
-	"@ai-sdk/google-vertex/anthropic": "google-vertex-anthropic",
-	"@ai-sdk/amazon-bedrock":          "bedrock",
-	"@ai-sdk/azure":                   "azure",
-	"@openrouter/ai-sdk-provider":     "openrouter",
-	"@ai-sdk/vercel":                  "vercel",
-	"@ai-sdk/openai-compatible":       "openaicompat",
+// wireProtocol identifies which LLM API protocol an npm package speaks.
+// Fantasy implements three native protocols (openai, anthropic, google);
+// everything else in its providers/ tree is a thin wrapper around one of
+// them with a pre-baked default URL or auth scheme.
+type wireProtocol int
+
+const (
+	wireUnknown wireProtocol = iota
+	wireOpenAI
+	wireAnthropic
+	wireGoogle
+)
+
+// npmToWireProtocol maps npm package names from models.dev to the wire
+// protocol they speak. Provider-specific bundles (azure, bedrock, vercel,
+// openrouter, google-vertex, google-vertex-anthropic) are intentionally
+// absent — they have native top-level cases in CreateProvider and never
+// reach the auto-router. Providers not in this map but with an api URL
+// are auto-routed through the OpenAI-compatible wire.
+var npmToWireProtocol = map[string]wireProtocol{
+	"@ai-sdk/openai":            wireOpenAI,
+	"@ai-sdk/openai-compatible": wireOpenAI,
+	"@ai-sdk/anthropic":         wireAnthropic,
+	"@ai-sdk/google":            wireGoogle,
 }

internal/models/providers.go

@@ -9,7 +9,9 @@ import (
 	"io"
 	"maps"
 	"net/http"
+	"net/url"
 	"os"
+	"regexp"
 	"strings"
 	"time"
 
@@ -25,6 +27,7 @@ import (
 	openaisdk "github.com/charmbracelet/openai-go"
 
 	"github.com/mark3labs/kit/internal/auth"
+	"github.com/spf13/viper"
 )
 
 const (
@@ -164,6 +167,13 @@ type ProviderConfig struct {
 	ThinkingLevel    ThinkingLevel
 	DisableCaching   bool // Opt-out: set to true to disable automatic prompt caching
 
+	// ConfigStore is the per-instance configuration store used to resolve
+	// "explicitly set" precedence checks (isExplicitlySet), per-model
+	// settings, and right-sizing. When nil, the process-global viper store is
+	// used. Threading a per-Kit store here keeps generation-parameter
+	// precedence isolated between Kit instances in the same process.
+	ConfigStore *viper.Viper
+
 	// ProgressReaderFunc, when set, wraps an io.Reader with progress display
 	// for long operations like Ollama model pulls. The returned io.ReadCloser
 	// must be closed when done. When nil, the raw reader is consumed directly
@@ -212,8 +222,10 @@ func ParseModelString(modelString string) (provider, model string, err error) {
 //
 // Native providers: anthropic, openai, google, ollama, azure, google-vertex-anthropic,
 // openrouter, bedrock, vercel.
-// Any provider in models.dev with an api URL or openai-compatible npm package
-// is auto-routed through fantasy's openaicompat provider.
+// Any other provider in models.dev is auto-routed by wire protocol: its npm
+// package (or per-model override) selects the OpenAI, Anthropic, or Google
+// transport, using the provider's api URL as the base. Providers with an api
+// URL but an unrecognized npm package fall back to the OpenAI-compatible wire.
 func CreateProvider(ctx context.Context, config *ProviderConfig) (*ProviderResult, error) {
 	provider, modelName, err := ParseModelString(config.ModelString)
 	if err != nil {
@@ -327,43 +339,62 @@ func CreateProvider(ctx context.Context, config *ProviderConfig) (*ProviderResul
 
 // autoRouteProvider attempts to create a provider by looking up its npm package
 // in the models.dev database and routing through the appropriate fantasy provider.
-// For openai-compatible providers, it uses the api URL from models.dev.
-// Models may have a provider override that specifies a different npm package than
-// the provider's default (e.g., opencode's claude-opus-4-6 uses @ai-sdk/anthropic).
+// It routes on wire protocol (openai, anthropic, google) rather than per-npm
+// provider name: fantasy implements three native wire protocols, and every other
+// entry in its providers/ tree is a thin wrapper around one of them. Using the
+// provider's api URL from models.dev as the base URL, any proxy that re-flavors
+// one of these protocols (e.g. opencode's Gemini routes) Just Works.
+//
+// Models may carry a provider override that specifies a different npm package
+// than the provider's default (e.g. opencode's claude-* uses @ai-sdk/anthropic
+// and its gemini-* uses @ai-sdk/google), which is resolved first.
 func autoRouteProvider(ctx context.Context, config *ProviderConfig, provider, modelName string, registry *ModelsRegistry) (*ProviderResult, error) {
 	providerInfo := registry.GetProviderInfo(provider)
 	if providerInfo == nil {
 		return nil, fmt.Errorf("unsupported provider: %s (not found in model database)", provider)
 	}
 
-	// Check for model-specific provider override
+	// Resolve npm: per-model override > provider default.
 	npmPackage := providerInfo.NPM
 	if modelInfo := registry.LookupModel(provider, modelName); modelInfo != nil && modelInfo.ProviderNPM != "" {
 		npmPackage = modelInfo.ProviderNPM
 	}
 
-	// Determine the LLM provider for this npm package
-	llmProvider := npmToLLMProvider[npmPackage]
-	if llmProvider == "" && providerInfo.API != "" {
-		// Unknown npm but has API URL → route through openaicompat
-		llmProvider = "openaicompat"
+	wire, known := npmToWireProtocol[npmPackage]
+	if !known {
+		// Unknown npm but the provider has an API URL → assume OpenAI-compatible.
+		// (Preserves the long-standing "any provider in models.dev with an api URL
+		// is auto-routed through openaicompat" behaviour.)
+		if providerInfo.API == "" {
+			return nil, fmt.Errorf(
+				"cannot auto-route provider %s: npm package %q has no known wire protocol "+
+					"and the registry has no API URL (use --provider-url to override)",
+				provider, npmPackage,
+			)
+		}
+		wire = wireOpenAI
 	}
 
-	switch llmProvider {
-	case "openaicompat":
+	// All three wires use the provider's API URL from models.dev as the base.
+	if config.ProviderURL == "" && providerInfo.API != "" {
+		config.ProviderURL = providerInfo.API
+	}
+
+	switch wire {
+	case wireOpenAI:
+		// The native OpenAI SDK package (@ai-sdk/openai) speaks the Responses
+		// API; openai-compatible proxies (and unknown-npm fallbacks) use the
+		// chat-completions wire via fantasy's openaicompat provider.
+		if npmPackage == "@ai-sdk/openai" {
+			return createAutoRoutedOpenAIProvider(ctx, config, modelName, providerInfo)
+		}
 		return createAutoRoutedOpenAICompatProvider(ctx, config, modelName, providerInfo)
-	case "anthropic":
-		if config.ProviderURL == "" && providerInfo.API != "" {
-			config.ProviderURL = providerInfo.API
-		}
+	case wireAnthropic:
 		return createAutoRoutedAnthropicProvider(ctx, config, modelName, providerInfo)
-	case "openai":
-		if config.ProviderURL == "" && providerInfo.API != "" {
-			config.ProviderURL = providerInfo.API
-		}
-		return createAutoRoutedOpenAIProvider(ctx, config, modelName, providerInfo)
+	case wireGoogle:
+		return createAutoRoutedGoogleProvider(ctx, config, modelName, providerInfo)
 	default:
-		return nil, fmt.Errorf("unsupported provider: %s (npm: %s has no LLM provider mapping)", provider, npmPackage)
+		return nil, fmt.Errorf("internal error: unknown wire protocol for provider %s (npm: %s)", provider, npmPackage)
 	}
 }
 
@@ -480,6 +511,115 @@ func createAutoRoutedOpenAIProvider(ctx context.Context, config *ProviderConfig,
 	return &ProviderResult{Model: model, ProviderOptions: providerOpts}, nil
 }
 
+// createAutoRoutedGoogleProvider creates a Google (Gemini) provider for
+// third-party providers that expose a Gemini-compatible API (e.g. opencode's
+// Gemini routes, which carry an @ai-sdk/google per-model override).
+//
+// The underlying genai SDK always injects its own API version segment
+// ("v1beta") between the base URL and the resource path. When the proxy's
+// base URL from models.dev already carries a version segment (e.g. opencode's
+// https://opencode.ai/zen/v1), that produces a doubled ".../v1/v1beta/..."
+// path that the proxy rejects. In that case we install a transport that
+// strips the injected segment so the proxy's own version is used.
+func createAutoRoutedGoogleProvider(ctx context.Context, config *ProviderConfig, modelName string, info *ProviderInfo) (*ProviderResult, error) {
+	apiKey := resolveAPIKey(config.ProviderAPIKey, info.Env)
+	if apiKey == "" {
+		return nil, fmt.Errorf("%s API key not provided. Use --provider-api-key or set %s",
+			info.Name, strings.Join(info.Env, " / "))
+	}
+
+	opts := []google.Option{
+		google.WithGeminiAPIKey(apiKey),
+		google.WithName(info.ID),
+	}
+
+	if config.ProviderURL != "" {
+		opts = append(opts, google.WithBaseURL(config.ProviderURL))
+	}
+
+	// Decide whether the genai-injected version segment needs stripping.
+	var httpClient *http.Client
+	if basePath := versionedBasePath(config.ProviderURL); basePath != "" {
+		httpClient = newGeminiProxyHTTPClient(basePath, config.TLSSkipVerify)
+	} else if config.TLSSkipVerify {
+		httpClient = createHTTPClientWithTLSConfig(true)
+	}
+	if httpClient != nil {
+		opts = append(opts, google.WithHTTPClient(httpClient))
+	}
+
+	p, err := google.New(opts...)
+	if err != nil {
+		return nil, fmt.Errorf("failed to create %s provider: %w", info.Name, err)
+	}
+
+	model, err := p.LanguageModel(ctx, modelName)
+	if err != nil {
+		return nil, fmt.Errorf("failed to create %s model: %w", info.Name, err)
+	}
+
+	return &ProviderResult{Model: model}, nil
+}
+
+// versionSegmentRe matches a trailing API version segment in a URL path,
+// e.g. "/v1", "/v1beta", "/v1beta1", "/v2alpha".
+var versionSegmentRe = regexp.MustCompile(`/v\d+(?:beta\d*|alpha\d*)?$`)
+
+// versionedBasePath returns the path component of rawURL when that path ends
+// with an API version segment (e.g. opencode's ".../zen/v1" → "/zen/v1").
+// It returns "" when rawURL is empty, unparseable, or has no version suffix
+// — in which case the genai SDK's default version injection is correct and
+// no rewriting is needed.
+func versionedBasePath(rawURL string) string {
+	if rawURL == "" {
+		return ""
+	}
+	u, err := url.Parse(rawURL)
+	if err != nil {
+		return ""
+	}
+	path := strings.TrimSuffix(u.Path, "/")
+	if versionSegmentRe.MatchString(path) {
+		return path
+	}
+	return ""
+}
+
+// newGeminiProxyHTTPClient builds an HTTP client whose transport strips the
+// genai-injected version segment ("v1beta"/"v1beta1") that directly follows
+// basePath, collapsing "{basePath}/v1beta/..." back to "{basePath}/...".
+func newGeminiProxyHTTPClient(basePath string, skipVerify bool) *http.Client {
+	var base http.RoundTripper
+	if skipVerify {
+		base = &http.Transport{TLSClientConfig: &tls.Config{InsecureSkipVerify: true}}
+	} else {
+		base = http.DefaultTransport
+	}
+	return &http.Client{
+		Transport: &geminiProxyTransport{base: base, basePath: basePath},
+	}
+}
+
+// geminiProxyTransport removes the redundant API version segment that the
+// genai SDK injects after a proxy base URL that already carries its own
+// version segment.
+type geminiProxyTransport struct {
+	base     http.RoundTripper
+	basePath string
+}
+
+func (t *geminiProxyTransport) RoundTrip(req *http.Request) (*http.Response, error) {
+	for _, injected := range []string{"/v1beta1", "/v1beta"} {
+		prefix := t.basePath + injected + "/"
+		if strings.HasPrefix(req.URL.Path, prefix) {
+			newReq := req.Clone(req.Context())
+			newReq.URL.Path = t.basePath + strings.TrimPrefix(req.URL.Path, t.basePath+injected)
+			return t.base.RoundTrip(newReq)
+		}
+	}
+	return t.base.RoundTrip(req)
+}
+
 // resolveAPIKey returns the first non-empty API key from the explicit key
 // or the environment variables.
 func resolveAPIKey(explicitKey string, envVars []string) string {
@@ -530,7 +670,7 @@ func rightSizeMaxTokens(config *ProviderConfig, modelInfo *ModelInfo) {
 	if modelInfo == nil || modelInfo.Limit.Output <= 0 {
 		return
 	}
-	if isExplicitlySet("max-tokens") {
+	if isExplicitlySet(config.ConfigStore, "max-tokens") {
 		return
 	}
 	target := min(modelInfo.Limit.Output, defaultRightSizeCap)

internal/models/registry.go

@@ -404,8 +404,8 @@ func isProviderLLMSupported(providerID string, info *ProviderInfo) bool {
 		return true
 	}
 
-	// Check if npm maps to an LLM provider
-	if _, ok := npmToLLMProvider[info.NPM]; ok {
+	// Check if npm maps to a known wire protocol
+	if _, ok := npmToWireProtocol[info.NPM]; ok {
 		return true
 	}
 

internal/prompts/loader.go

@@ -36,15 +36,17 @@ type Diagnostic struct {
 }
 
 // LoadAll discovers and loads all prompt templates from standard locations
-// and any extra paths. Templates are loaded in order of precedence (lowest
-// to highest), with later templates overriding earlier ones of the same name.
+// and any extra paths. Templates are loaded in order of precedence (highest
+// to lowest); the first source to define a given name wins, later definitions
+// of the same name are dropped with a diagnostic.
 //
 // Discovery paths searched in order:
 //  1. Default templates (if IncludeDefaults)
-//  2. ~/.kit/prompts/ (global user templates)
-//  3. .kit/prompts/ (project-local templates)
-//  4. ConfigPaths (from configuration)
-//  5. ExtraPaths (explicit paths, highest precedence)
+//  2. ~/.kit/prompts/                          (legacy global)
+//  3. $XDG_CONFIG_HOME/kit/prompts/            (XDG global, default ~/.config/kit/prompts/)
+//  4. <cwd>/.kit/prompts/                      (project-local templates)
+//  5. ConfigPaths                              (from configuration)
+//  6. ExtraPaths                               (explicit paths, lowest precedence)
 func LoadAll(opts LoadOptions) ([]*PromptTemplate, []Diagnostic, error) {
 	if opts.Cwd == "" {
 		opts.Cwd, _ = os.Getwd()
@@ -88,13 +90,21 @@ func LoadAll(opts LoadOptions) ([]*PromptTemplate, []Diagnostic, error) {
 		addTemplates(defaults, "default")
 	}
 
-	// 2. Global user templates: ~/.kit/prompts/
-	globalDir := filepath.Join(opts.HomeDir, ".kit", "prompts")
-	if templates, err := LoadFromDir(globalDir); err == nil {
+	// 2. Legacy global user templates: ~/.kit/prompts/
+	legacyGlobalDir := filepath.Join(opts.HomeDir, ".kit", "prompts")
+	if templates, err := LoadFromDir(legacyGlobalDir); err == nil {
 		addTemplates(templates, "global")
 	}
 
-	// 3. Project-local templates: .kit/prompts/
+	// 3. XDG global user templates: $XDG_CONFIG_HOME/kit/prompts/
+	//    Default: ~/.config/kit/prompts/. Aligns with extensions and skills.
+	if xdgDir := GlobalDir(); xdgDir != "" && xdgDir != legacyGlobalDir {
+		if templates, err := LoadFromDir(xdgDir); err == nil {
+			addTemplates(templates, "global")
+		}
+	}
+
+	// 4. Project-local templates: .kit/prompts/
 	localDir := filepath.Join(opts.Cwd, ".kit", "prompts")
 	if templates, err := LoadFromDir(localDir); err == nil {
 		addTemplates(templates, "local")
@@ -186,3 +196,22 @@ func loadDefaultTemplates() []*PromptTemplate {
 	// For now, return an empty slice - users can define their own templates
 	return nil
 }
+
+// GlobalDir returns the XDG-aligned global prompts directory, respecting
+// $XDG_CONFIG_HOME. Defaults to ~/.config/kit/prompts/. Returns an empty
+// string if the user's home directory cannot be resolved.
+//
+// This is the canonical location for user-wide prompt templates and aligns
+// with the discovery paths used for extensions ($XDG_CONFIG_HOME/kit/extensions/)
+// and skills ($XDG_CONFIG_HOME/kit/skills/).
+func GlobalDir() string {
+	base := os.Getenv("XDG_CONFIG_HOME")
+	if base == "" {
+		home, err := os.UserHomeDir()
+		if err != nil {
+			return ""
+		}
+		base = filepath.Join(home, ".config")
+	}
+	return filepath.Join(base, "kit", "prompts")
+}

internal/session/compaction_test.go

@@ -129,26 +129,35 @@ func TestCompactionWithNewMessagesAfterCompaction(t *testing.T) {
 	msg4 := message.Message{Role: message.RoleAssistant, Parts: []message.ContentPart{message.TextContent{Text: "Message 4 - after compaction"}}}
 	_, _ = tm.AppendMessage(msg4)
 
-	// BuildContext should return: [summary] + [M4 (new after compaction)] + [M3 (kept)]
+	// BuildContext should return: [summary] + [M3 (kept)] + [M4 (new after compaction)]
+	// Kept messages must appear BEFORE post-compaction messages so the LLM
+	// sees the conversation in chronological order. Otherwise the latest
+	// post-compaction user message would be followed by an older kept user
+	// message, breaking user/assistant alternation and causing the model to
+	// respond as if the post-compaction turn never happened.
 	messages, _, _ := tm.BuildContext()
 	if len(messages) != 3 {
-		t.Fatalf("expected 3 messages (summary + M4 + M3), got %d: %+v", len(messages), messages)
+		t.Fatalf("expected 3 messages (summary + M3 + M4), got %d: %+v", len(messages), messages)
 	}
 
-	// Verify order: summary, M4 (new), M3 (kept)
+	// Verify order: summary, M3 (kept), M4 (new)
 	if messages[0].Role != fantasy.MessageRoleSystem {
 		t.Errorf("first message should be summary, got %s", messages[0].Role)
 	}
-	if messages[1].Role != fantasy.MessageRoleAssistant {
-		t.Errorf("second message should be assistant (M4), got %s", messages[1].Role)
+	if messages[1].Role != fantasy.MessageRoleUser {
+		t.Errorf("second message should be user (M3 kept), got %s", messages[1].Role)
 	}
-	m4Text := messages[1].Content[0].(fantasy.TextPart).Text
+	m3Text := messages[1].Content[0].(fantasy.TextPart).Text
+	if m3Text != "Message 3 - kept" {
+		t.Errorf("unexpected M3 text: %s", m3Text)
+	}
+	if messages[2].Role != fantasy.MessageRoleAssistant {
+		t.Errorf("third message should be assistant (M4 post-compact), got %s", messages[2].Role)
+	}
+	m4Text := messages[2].Content[0].(fantasy.TextPart).Text
 	if m4Text != "Message 4 - after compaction" {
 		t.Errorf("unexpected M4 text: %s", m4Text)
 	}
-	if messages[2].Role != fantasy.MessageRoleUser {
-		t.Errorf("third message should be user (M3), got %s", messages[2].Role)
-	}
 
 	// Verify that M1 is NOT in the context
 	for i, msg := range messages {

internal/session/store.go

@@ -6,8 +6,10 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime"
 	"sort"
 	"strings"
+	"sync"
 	"time"
 )
 
@@ -97,6 +99,11 @@ func ListAllSessions() ([]SessionInfo, error) {
 
 // listSessionsInDir reads all .jsonl files in a directory and extracts session info.
 // Empty sessions (no messages) are automatically cleaned up and not returned.
+//
+// Per-file extraction is parallelized across a small worker pool because each
+// file requires a full JSONL scan to compute MessageCount and FirstMessage —
+// for users with many sessions this is the dominant cost of opening the
+// session picker.
 func listSessionsInDir(dir string) ([]SessionInfo, error) {
 	if _, err := os.Stat(dir); os.IsNotExist(err) {
 		return nil, nil
@@ -107,20 +114,47 @@ func listSessionsInDir(dir string) ([]SessionInfo, error) {
 		return nil, fmt.Errorf("failed to read directory %s: %w", dir, err)
 	}
 
-	var sessions []SessionInfo
+	// Collect candidate paths first so we can parallelize the heavy work.
+	paths := make([]string, 0, len(entries))
 	for _, entry := range entries {
 		if entry.IsDir() || !strings.HasSuffix(entry.Name(), ".jsonl") {
 			continue
 		}
+		paths = append(paths, filepath.Join(dir, entry.Name()))
+	}
 
-		path := filepath.Join(dir, entry.Name())
-		info, err := extractSessionInfo(path)
-		if err != nil {
-			continue // skip malformed session files
+	results := make([]*SessionInfo, len(paths))
+
+	// Worker pool sized to GOMAXPROCS, capped to avoid thrashing for tiny lists.
+	workers := max(min(runtime.GOMAXPROCS(0), len(paths)), 1)
+
+	var wg sync.WaitGroup
+	jobs := make(chan int, len(paths))
+	for range workers {
+		wg.Go(func() {
+			for i := range jobs {
+				info, err := extractSessionInfo(paths[i])
+				if err != nil {
+					continue // skip malformed session files
+				}
+				results[i] = info
+			}
+		})
+	}
+	for i := range paths {
+		jobs <- i
+	}
+	close(jobs)
+	wg.Wait()
+
+	sessions := make([]SessionInfo, 0, len(results))
+	for i, info := range results {
+		if info == nil {
+			continue
 		}
-		// Clean up and skip empty sessions (no messages)
+		// Clean up and skip empty sessions (no messages).
 		if info.MessageCount == 0 {
-			_ = os.Remove(path)
+			_ = os.Remove(paths[i])
 			continue
 		}
 		sessions = append(sessions, *info)

internal/session/tree_manager.go

@@ -755,9 +755,17 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 		}
 	}
 
-	// If there is a compaction, inject the summary first and collect
-	// the kept messages starting from FirstKeptEntryID (since the
-	// compaction entry's parent chain doesn't include them).
+	// If there is a compaction, inject the summary first, then the
+	// preserved "kept" messages (chronologically before the compaction),
+	// then the post-compaction messages (chronologically after).
+	//
+	// Order matters: the kept messages must come BEFORE the post-compaction
+	// branch so the LLM sees the conversation in chronological order. If the
+	// kept messages were appended last, the latest user message in the
+	// current branch would be followed by an older kept user message,
+	// breaking the strict user/assistant alternation that providers expect
+	// and causing the model to respond as if the previous turn never
+	// happened.
 	if lastCompaction != nil {
 		messages = append(messages, fantasy.Message{
 			Role: fantasy.MessageRoleSystem,
@@ -768,49 +776,10 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 			},
 		})
 
-		// Collect entries from the compaction entry itself (at compactionIndex)
-		// and any entries before it in the branch (newer messages).
-		for i := compactionIndex; i < len(branch); i++ {
-			entry := branch[i]
-			switch e := entry.(type) {
-			case *MessageEntry:
-				msg, err := e.ToMessage()
-				if err != nil {
-					continue // skip malformed entries
-				}
-				msgs := msg.ToLLMMessages()
-				messages = append(messages, msgs...)
-
-			case *BranchSummaryEntry:
-				// Convert branch summary to a user message for context.
-				if e.Summary != "" {
-					messages = append(messages, fantasy.Message{
-						Role: fantasy.MessageRoleUser,
-						Content: []fantasy.MessagePart{
-							fantasy.TextPart{
-								Text: fmt.Sprintf("[Branch context: %s]", e.Summary),
-							},
-						},
-					})
-				}
-
-			case *ModelChangeEntry:
-				provider = e.Provider
-				modelID = e.ModelID
-
-			case *CompactionEntry:
-				// Already handled above (summary injected).
-				continue
-			}
-		}
-
-		// Now collect the kept messages starting from FirstKeptEntryID.
-		// These are not in the current branch because the compaction entry
-		// is parented to the first kept entry's parent, not the first kept entry.
-		// We iterate through entries in order (not using getBranchLocked) to avoid
-		// walking back to old compacted messages.
-		// We stop when we reach the compaction entry to avoid double-counting
-		// messages that were added after the compaction.
+		// Step 1: collect the kept messages starting from FirstKeptEntryID.
+		// These are not on the current branch (the compaction entry is a
+		// new root with no parent), so we iterate tm.entries in append order
+		// and stop when we reach the compaction entry itself.
 		if lastCompaction.FirstKeptEntryID != "" {
 			found := false
 			for _, entry := range tm.entries {
@@ -825,13 +794,12 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 					}
 				}
 
-				// Stop when we reach the compaction entry itself.
-				// Messages after the compaction are collected from the branch walk above.
+				// Stop when we reach the compaction entry itself; messages
+				// after it are collected from the branch walk below.
 				if entryID == lastCompaction.ID {
 					break
 				}
 
-				// Process this kept entry.
 				switch e := entry.(type) {
 				case *MessageEntry:
 					msg, err := e.ToMessage()
@@ -860,6 +828,42 @@ func (tm *TreeManager) BuildContext() (messages []fantasy.Message, provider stri
 			}
 		}
 
+		// Step 2: collect entries on the current branch after the compaction
+		// entry (these are post-compaction messages). The compaction entry
+		// itself is skipped — its summary was already injected above.
+		for i := compactionIndex; i < len(branch); i++ {
+			entry := branch[i]
+			switch e := entry.(type) {
+			case *MessageEntry:
+				msg, err := e.ToMessage()
+				if err != nil {
+					continue
+				}
+				msgs := msg.ToLLMMessages()
+				messages = append(messages, msgs...)
+
+			case *BranchSummaryEntry:
+				if e.Summary != "" {
+					messages = append(messages, fantasy.Message{
+						Role: fantasy.MessageRoleUser,
+						Content: []fantasy.MessagePart{
+							fantasy.TextPart{
+								Text: fmt.Sprintf("[Branch context: %s]", e.Summary),
+							},
+						},
+					})
+				}
+
+			case *ModelChangeEntry:
+				provider = e.Provider
+				modelID = e.ModelID
+
+			case *CompactionEntry:
+				// Summary already injected above.
+				continue
+			}
+		}
+
 		return messages, provider, modelID
 	}
 
@@ -1030,44 +1034,22 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 
 	var ids []string
 
-	// If there's a compaction, we need to collect IDs from:
-	// 1. Entries after the compaction entry in the branch (newer messages)
-	// 2. Entries from FirstKeptEntryID onwards (kept messages)
+	// If there's a compaction, we collect IDs in the same order as
+	// BuildContext: [summary placeholder, kept messages, post-compaction
+	// messages]. This ordering must stay in sync with BuildContext so a
+	// cut-point index can be mapped back to the correct entry ID.
 	if lastCompaction != nil {
 		// Placeholder for the summary system message (no entry ID).
 		ids = append(ids, "")
 
-		// Collect IDs from entries after the compaction entry (newer messages).
-		for i := compactionIndex + 1; i < len(branch); i++ {
-			entry := branch[i]
-			switch e := entry.(type) {
-			case *MessageEntry:
-				msg, err := e.ToMessage()
-				if err != nil {
-					continue
-				}
-				msgs := msg.ToLLMMessages()
-				for range msgs {
-					ids = append(ids, e.ID)
-				}
-
-			case *BranchSummaryEntry:
-				if e.Summary != "" {
-					ids = append(ids, e.ID)
-				}
-			}
-		}
-
-		// Collect IDs from the kept messages starting at FirstKeptEntryID.
-		// We iterate through entries in order (not using getBranchLocked) to avoid
-		// walking back to old compacted messages.
-		// We stop when we reach the compaction entry to avoid double-counting.
+		// Step 1: IDs of the kept messages starting at FirstKeptEntryID.
+		// Iterate tm.entries in append order and stop at the compaction
+		// entry to avoid double-counting post-compaction messages.
 		if lastCompaction.FirstKeptEntryID != "" {
 			found := false
 			for _, entry := range tm.entries {
 				entryID := tm.EntryID(entry)
 
-				// Skip entries until we reach the first kept entry.
 				if !found {
 					if entryID == lastCompaction.FirstKeptEntryID {
 						found = true
@@ -1076,7 +1058,6 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 					}
 				}
 
-				// Stop when we reach the compaction entry itself.
 				if entryID == lastCompaction.ID {
 					break
 				}
@@ -1100,6 +1081,28 @@ func (tm *TreeManager) GetContextEntryIDs() []string {
 			}
 		}
 
+		// Step 2: IDs of entries after the compaction entry on the current
+		// branch (post-compaction messages).
+		for i := compactionIndex + 1; i < len(branch); i++ {
+			entry := branch[i]
+			switch e := entry.(type) {
+			case *MessageEntry:
+				msg, err := e.ToMessage()
+				if err != nil {
+					continue
+				}
+				msgs := msg.ToLLMMessages()
+				for range msgs {
+					ids = append(ids, e.ID)
+				}
+
+			case *BranchSummaryEntry:
+				if e.Summary != "" {
+					ids = append(ids, e.ID)
+				}
+			}
+		}
+
 		return ids
 	}
 

internal/tools/connection_pool.go

@@ -345,49 +345,70 @@ func (p *MCPConnectionPool) createStdioClient(ctx context.Context, serverConfig
 	return stdioClient, nil
 }
 
-// createSSEClient creates an SSE client
+// parseHeaders parses "Key: Value" header strings into a map.
+func parseHeaders(raw []string) map[string]string {
+	if len(raw) == 0 {
+		return nil
+	}
+	headers := make(map[string]string)
+	for _, header := range raw {
+		parts := strings.SplitN(header, ":", 2)
+		if len(parts) == 2 {
+			key := strings.TrimSpace(parts[0])
+			value := strings.TrimSpace(parts[1])
+			headers[key] = value
+		}
+	}
+	if len(headers) == 0 {
+		return nil
+	}
+	return headers
+}
+
+// buildOAuthConfig constructs a transport.OAuthConfig from the server config
+// and the pool's OAuth flow. Returns nil if OAuth is not applicable.
+func (p *MCPConnectionPool) buildOAuthConfig(serverConfig config.MCPServerConfig) (*transport.OAuthConfig, error) {
+	if p.oauthFlow == nil || serverConfig.NoOAuth {
+		return nil, nil
+	}
+	tokenStore, err := p.createTokenStore(serverConfig.URL)
+	if err != nil {
+		return nil, fmt.Errorf("failed to create token store: %w", err)
+	}
+	cfg := &transport.OAuthConfig{
+		RedirectURI: p.oauthFlow.handler.RedirectURI(),
+		PKCEEnabled: true,
+		TokenStore:  tokenStore,
+	}
+	if serverConfig.OAuthClientID != "" {
+		cfg.ClientID = serverConfig.OAuthClientID
+	}
+	if serverConfig.OAuthClientSecret != "" {
+		cfg.ClientSecret = serverConfig.OAuthClientSecret
+	}
+	if len(serverConfig.OAuthScopes) > 0 {
+		cfg.Scopes = serverConfig.OAuthScopes
+	}
+	return cfg, nil
+}
+
 func (p *MCPConnectionPool) createSSEClient(ctx context.Context, serverConfig config.MCPServerConfig) (client.MCPClient, error) {
 	var options []transport.ClientOption
 
-	if len(serverConfig.Headers) > 0 {
-		headers := make(map[string]string)
-		for _, header := range serverConfig.Headers {
-			parts := strings.SplitN(header, ":", 2)
-			if len(parts) == 2 {
-				key := strings.TrimSpace(parts[0])
-				value := strings.TrimSpace(parts[1])
-				headers[key] = value
-			}
-		}
-		if len(headers) > 0 {
-			options = append(options, transport.WithHeaders(headers))
-		}
+	if headers := parseHeaders(serverConfig.Headers); headers != nil {
+		options = append(options, transport.WithHeaders(headers))
 	}
 
 	// Enable OAuth for remote transports when an auth handler is configured
 	// and the server hasn't opted out via NoOAuth. Public MCP servers (e.g.
 	// PubMed) set NoOAuth to skip dynamic client registration and token
 	// exchange, which would otherwise fail with a 404.
-	if p.oauthFlow != nil && !serverConfig.NoOAuth {
-		tokenStore, tsErr := p.createTokenStore(serverConfig.URL)
-		if tsErr != nil {
-			return nil, fmt.Errorf("failed to create token store: %w", tsErr)
-		}
-		oauthCfg := transport.OAuthConfig{
-			RedirectURI: p.oauthFlow.handler.RedirectURI(),
-			PKCEEnabled: true,
-			TokenStore:  tokenStore,
-		}
-		if serverConfig.OAuthClientID != "" {
-			oauthCfg.ClientID = serverConfig.OAuthClientID
-		}
-		if serverConfig.OAuthClientSecret != "" {
-			oauthCfg.ClientSecret = serverConfig.OAuthClientSecret
-		}
-		if len(serverConfig.OAuthScopes) > 0 {
-			oauthCfg.Scopes = serverConfig.OAuthScopes
-		}
-		options = append(options, transport.WithOAuth(oauthCfg))
+	oauthCfg, err := p.buildOAuthConfig(serverConfig)
+	if err != nil {
+		return nil, err
+	}
+	if oauthCfg != nil {
+		options = append(options, transport.WithOAuth(*oauthCfg))
 	}
 
 	sseClient, err := client.NewSSEMCPClient(serverConfig.URL, options...)
@@ -406,43 +427,18 @@ func (p *MCPConnectionPool) createSSEClient(ctx context.Context, serverConfig co
 func (p *MCPConnectionPool) createStreamableClient(ctx context.Context, serverConfig config.MCPServerConfig) (client.MCPClient, error) {
 	var options []transport.StreamableHTTPCOption
 
-	if len(serverConfig.Headers) > 0 {
-		headers := make(map[string]string)
-		for _, header := range serverConfig.Headers {
-			parts := strings.SplitN(header, ":", 2)
-			if len(parts) == 2 {
-				key := strings.TrimSpace(parts[0])
-				value := strings.TrimSpace(parts[1])
-				headers[key] = value
-			}
-		}
-		if len(headers) > 0 {
-			options = append(options, transport.WithHTTPHeaders(headers))
-		}
+	if headers := parseHeaders(serverConfig.Headers); headers != nil {
+		options = append(options, transport.WithHTTPHeaders(headers))
 	}
 
 	// Enable OAuth for remote transports when an auth handler is configured
 	// and the server hasn't opted out via NoOAuth.
-	if p.oauthFlow != nil && !serverConfig.NoOAuth {
-		tokenStore, tsErr := p.createTokenStore(serverConfig.URL)
-		if tsErr != nil {
-			return nil, fmt.Errorf("failed to create token store: %w", tsErr)
-		}
-		oauthCfg := transport.OAuthConfig{
-			RedirectURI: p.oauthFlow.handler.RedirectURI(),
-			PKCEEnabled: true,
-			TokenStore:  tokenStore,
-		}
-		if serverConfig.OAuthClientID != "" {
-			oauthCfg.ClientID = serverConfig.OAuthClientID
-		}
-		if serverConfig.OAuthClientSecret != "" {
-			oauthCfg.ClientSecret = serverConfig.OAuthClientSecret
-		}
-		if len(serverConfig.OAuthScopes) > 0 {
-			oauthCfg.Scopes = serverConfig.OAuthScopes
-		}
-		options = append(options, transport.WithHTTPOAuth(oauthCfg))
+	oauthCfg, err := p.buildOAuthConfig(serverConfig)
+	if err != nil {
+		return nil, err
+	}
+	if oauthCfg != nil {
+		options = append(options, transport.WithHTTPOAuth(*oauthCfg))
 	}
 
 	streamableClient, err := client.NewStreamableHttpClient(serverConfig.URL, options...)

internal/ui/commands/commands.go

@@ -161,6 +161,12 @@ var SlashCommands = []SlashCommand{
 		Category:    "Navigation",
 		Aliases:     []string{"/r"},
 	},
+	{
+		Name:        "/copy",
+		Description: "Copy the last message to the system clipboard",
+		Category:    "System",
+		Aliases:     []string{"/cp"},
+	},
 	{
 		Name:        "/export",
 		Description: "Export session (JSONL by default, or /export path.jsonl)",

internal/ui/model.go

@@ -135,6 +135,15 @@ type SkillItem struct {
 	Description string // Short summary used in autocomplete and help.
 }
 
+// ExtensionItem holds display metadata about a loaded extension for the
+// startup [Extensions] section. Built by the CLI layer from the SDK's
+// []kit.ExtensionInfo.
+type ExtensionItem struct {
+	Name   string // Extension display name (filename without .go extension).
+	Path   string // Absolute path to the extension's .go file.
+	Source string // "project" or "user" (global).
+}
+
 // MCPPromptInfo describes an MCP prompt for display in the TUI (autocomplete,
 // help). This is a pure UI type — it carries no MCP client dependencies.
 type MCPPromptInfo struct {
@@ -364,6 +373,16 @@ type AppModelOptions struct {
 	// watcher detects changes. May be nil if skill hot-reload is not needed.
 	GetSkillItems func() []SkillItem
 
+	// ExtensionItems lists loaded extensions for the [Extensions] startup
+	// section. Each entry shows the filename of an extension that was
+	// discovered and loaded (global, project-local, or explicit).
+	ExtensionItems []ExtensionItem
+
+	// GetExtensionItems, if non-nil, returns the current extension items.
+	// Called on extension hot-reload to refresh the list. May be nil if no
+	// extensions are loaded.
+	GetExtensionItems func() []ExtensionItem
+
 	// MCPToolCount is the number of tools loaded from external MCP servers.
 	MCPToolCount int
 
@@ -608,6 +627,14 @@ type AppModel struct {
 	// skill list after content hot-reload. May be nil.
 	getSkillItems func() []SkillItem
 
+	// extensionItems lists loaded extensions for the [Extensions] startup
+	// section (filenames only).
+	extensionItems []ExtensionItem
+
+	// getExtensionItems returns the current extension items. Used to refresh
+	// the list after extension hot-reload. May be nil.
+	getExtensionItems func() []ExtensionItem
+
 	// mcpToolCount and extensionToolCount track tool counts by source for
 	// the startup info display.
 	mcpToolCount       int
@@ -861,6 +888,8 @@ func NewAppModel(appCtrl AppController, opts AppModelOptions) *AppModel {
 	m.contextPaths = opts.ContextPaths
 	m.skillItems = opts.SkillItems
 	m.getSkillItems = opts.GetSkillItems
+	m.extensionItems = opts.ExtensionItems
+	m.getExtensionItems = opts.GetExtensionItems
 	m.mcpToolCount = opts.MCPToolCount
 	m.extensionToolCount = opts.ExtensionToolCount
 	m.startupExtensionMessages = opts.StartupExtensionMessages
@@ -1029,8 +1058,21 @@ func (m *AppModel) AddStartupMessageToScrollList() {
 		pairs = append(pairs, [2]string{"Skills", strings.Join(names, ", ")})
 	}
 
-	// Extension tool count (only shown when > 0).
-	if m.extensionToolCount > 0 {
+	// Extensions — listed by filename. Each extension shows its basename
+	// without the .go suffix, matching the [Skills] section's style.
+	if len(m.extensionItems) > 0 {
+		names := make([]string, len(m.extensionItems))
+		for i, ei := range m.extensionItems {
+			names[i] = ei.Name
+		}
+		value := strings.Join(names, ", ")
+		if m.extensionToolCount > 0 {
+			value += fmt.Sprintf(" (%d tools)", m.extensionToolCount)
+		}
+		pairs = append(pairs, [2]string{"Extensions", value})
+	} else if m.extensionToolCount > 0 {
+		// Fallback: tool count only (extensions registered tools but the CLI
+		// did not provide ExtensionItems for some reason).
 		pairs = append(pairs, [2]string{"Extensions", fmt.Sprintf("%d tools", m.extensionToolCount)})
 	}
 
@@ -1266,7 +1308,11 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			m.scrollList.autoScroll = false
 		case tea.MouseWheelDown:
 			m.scrollList.ScrollBy(scrollLines)
-			if m.scrollList.AtBottom() {
+			// Only re-enable auto-scroll when the user is not actively
+			// selecting text. Otherwise a wheel-down during a drag-select
+			// would re-arm GotoBottom on the next stream chunk, shifting
+			// the highlighted row out from under the cursor.
+			if m.scrollList.AtBottom() && !m.scrollList.IsMouseDown() {
 				m.scrollList.autoScroll = true
 			}
 		}
@@ -1274,9 +1320,14 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 	// ── Mouse click selection (crush-style character-level) ──────────────────
 	case tea.MouseClickMsg:
 		if msg.Button == tea.MouseLeft {
-			// Calculate viewport-relative coordinates.
-			viewY := msg.Y - m.scrollbackYOffset
-			if viewY >= 0 && viewY < m.scrollList.height {
+			// Compute the scrollback origin from the current frame's layout
+			// rather than the stale cached value from the previous View().
+			// scrollbackYOffset/scrollList.height are only refreshed inside
+			// View() and lag behind any state change that resized the header
+			// (extension widgets, warning rows, etc.) since the last render.
+			yOff, vpHeight := m.currentScrollbackBounds()
+			viewY := msg.Y - yOff
+			if viewY >= 0 && viewY < vpHeight {
 				// Clear any previous selection on a new click.
 				// HandleMouseDown will set up new selection state.
 				if m.scrollList.HandleMouseDown(msg.X, viewY) {
@@ -1287,8 +1338,9 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 
 	// ── Mouse motion/drag for character-level selection ──────────────────────
 	case tea.MouseMotionMsg:
-		viewY := msg.Y - m.scrollbackYOffset
-		if viewY >= 0 && viewY < m.scrollList.height {
+		yOff, vpHeight := m.currentScrollbackBounds()
+		viewY := msg.Y - yOff
+		if viewY >= 0 && viewY < vpHeight {
 			m.scrollList.HandleMouseDrag(msg.X, viewY)
 		}
 
@@ -1618,10 +1670,16 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 
 	// ── Cancel timer expired ─────────────────────────────────────────────────
 	case uicore.CancelTimerExpiredMsg:
+		if m.canceling {
+			m.layoutDirty = true
+		}
 		m.canceling = false
 
 	// ── Ctrl+C reset timer expired ────────────────────────────────────────────
 	case uicore.CtrlCResetMsg:
+		if m.ctrlCPressedOnce {
+			m.layoutDirty = true
+		}
 		m.ctrlCPressedOnce = false
 
 	// ── Input submitted ──────────────────────────────────────────────────────
@@ -2343,6 +2401,7 @@ func (m *AppModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		if msg.err != nil {
 			m.printSystemMessage(fmt.Sprintf("Extension reload failed: %v", msg.err))
 		} else {
+			m.refreshExtensionItems()
 			m.printSystemMessage("Extensions reloaded.")
 		}
 
@@ -3110,6 +3169,8 @@ func (m *AppModel) handleSlashCommand(sc *commands.SlashCommand, args string) te
 		return m.handleResumeCommand()
 	case "/export":
 		return m.handleExportCommand(args)
+	case "/copy":
+		return m.handleCopyCommand()
 	case "/share":
 		return m.handleShareCommand()
 	case "/import":
@@ -3439,6 +3500,16 @@ func (m *AppModel) refreshSkillItems() {
 	}
 }
 
+// refreshExtensionItems reloads extension items from the provider callback
+// so the [Extensions] startup section reflects the current set after a
+// hot-reload. Called from the extReloadResultMsg handler.
+func (m *AppModel) refreshExtensionItems() {
+	if m.getExtensionItems == nil {
+		return
+	}
+	m.extensionItems = m.getExtensionItems()
+}
+
 // formatSkillDescription returns the autocomplete description for a skill,
 // prefixed with [project] or [user] so users can tell colliding names apart.
 func formatSkillDescription(s SkillItem) string {
@@ -3524,6 +3595,7 @@ func (m *AppModel) printHelpMessage() {
 		"**System:**\n" +
 		"- `/compact [instructions]`: Summarise older messages to free context space\n" +
 		"- `/clear`: Clear message history\n" +
+		"- `/copy`: Copy the last message to the system clipboard\n" +
 		"- `/export [path]`: Export session as JSONL\n" +
 		"- `/import <path.jsonl>`: Import session from JSONL file\n" +
 		"- `/reset-usage`: Reset usage statistics\n" +
@@ -3760,7 +3832,12 @@ func (m *AppModel) appendStreamingChunk(role, content string) {
 		}
 		// Auto-scroll to bottom if enabled (iteratr pattern)
 		// Don't call SetItems() - the slice reference hasn't changed
-		if m.scrollList != nil {
+		//
+		// CRITICAL: never scroll the viewport while the user is actively
+		// selecting text (mouse button held). Doing so shifts the
+		// highlighted content out from under the cursor and produces the
+		// off-by-N-row drift users see when copy-selecting during streaming.
+		if m.scrollList != nil && !m.scrollList.IsMouseDown() {
 			if m.scrollList.autoScroll {
 				m.scrollList.GotoBottom()
 			} else if m.scrollList.AtBottom() {
@@ -3788,6 +3865,36 @@ func (m *AppModel) appendStreamingChunk(role, content string) {
 	m.refreshContent()
 }
 
+// currentScrollbackBounds returns the live (yOffset, viewportHeight) for the
+// scrollback region, computed from the current state — not from the cached
+// values populated inside View().
+//
+// scrollbackYOffset and scrollList.height are refreshed once per render, so
+// any state change that resizes the header (extension widget toggles,
+// warning rows, queued messages, etc.) leaves the cached values one frame
+// stale. Mouse click handlers in Update() can then place the cursor on the
+// wrong line, producing the off-by-N-row drift seen during copy-selection.
+//
+// This recomputes the header height by rendering it (cheap — the renderer
+// returns "" when no extension header is set) and recomputes the viewport
+// height the same way distributeHeight() does, so both inputs to the
+// y → (item, line) mapping are always current.
+func (m *AppModel) currentScrollbackBounds() (yOffset, viewportHeight int) {
+	// Force a fresh layout if anything in Update() marked the state dirty;
+	// otherwise scrollList.height still reflects the previous frame.
+	if m.layoutDirty {
+		m.distributeHeight()
+		m.layoutDirty = false
+	}
+	if headerView := m.renderHeaderFooter(m.getHeader); headerView != "" {
+		yOffset = lipgloss.Height(headerView)
+	}
+	if m.scrollList != nil {
+		viewportHeight = m.scrollList.height
+	}
+	return yOffset, viewportHeight
+}
+
 // distributeHeight recalculates child component heights after a window resize,
 // queue change, widget update, or state transition, and propagates the computed
 // stream height to the StreamComponent.
@@ -3860,7 +3967,20 @@ func (m *AppModel) distributeHeight() {
 		headerFooterLines += lipgloss.Height(footerView)
 	}
 
-	streamHeight := max(m.height-separatorLines-widgetLines-headerFooterLines-queuedLines-inputLines-statusBarLines, 0)
+	// Account for transient warning rows that View() injects between the
+	// scrollback and the separator. These flags are toggled by ESC/Ctrl+C
+	// handlers; without subtracting them here the joined view exceeds
+	// m.height by one line per active warning and the bottom of the screen
+	// gets silently clipped — which in turn invalidates scrollbackYOffset.
+	var warningLines int
+	if m.canceling {
+		warningLines++
+	}
+	if m.ctrlCPressedOnce {
+		warningLines++
+	}
+
+	streamHeight := max(m.height-separatorLines-widgetLines-headerFooterLines-queuedLines-inputLines-statusBarLines-warningLines, 0)
 
 	// In alt screen mode, give the calculated height to ScrollList instead of stream.
 	// The stream component still exists but is embedded as the last item in scrollList.
@@ -4284,6 +4404,48 @@ func (m *AppModel) handleNameCommand(args string) tea.Cmd {
 	return nil
 }
 
+// handleCopyCommand copies the last user or assistant message to the system
+// clipboard. Skips transient system messages (e.g. /help output) so the user
+// gets the actual last conversational message.
+func (m *AppModel) handleCopyCommand() tea.Cmd {
+	if len(m.messages) == 0 {
+		m.printSystemMessage("No messages to copy.")
+		return nil
+	}
+
+	var (
+		text string
+		role string
+	)
+	for i := len(m.messages) - 1; i >= 0; i-- {
+		switch msg := m.messages[i].(type) {
+		case *TextMessageItem:
+			if msg.role == "user" || msg.role == "assistant" {
+				text = msg.content
+				role = msg.role
+			}
+		case *StreamingMessageItem:
+			if msg.role == "assistant" || msg.role == "reasoning" {
+				text = msg.content.String()
+				role = msg.role
+			}
+		}
+		if text != "" {
+			break
+		}
+	}
+
+	if strings.TrimSpace(text) == "" {
+		m.printSystemMessage("No copyable message found.")
+		return nil
+	}
+
+	m.printSystemMessage(fmt.Sprintf(
+		"Copied last %s message to clipboard (%d chars).", role, len(text),
+	))
+	return clipboard.CopyToClipboard(text)
+}
+
 // handleExportCommand exports the current session to a file.
 // Usage: /export          — copies the JSONL file to cwd with a descriptive name.
 //

internal/ui/scrolllist.go

@@ -60,10 +60,13 @@ func NewScrollList(width, height int) *ScrollList {
 }
 
 // SetItems replaces the items in the scroll list. If auto-scroll is enabled,
-// the viewport will scroll to the bottom to show the latest content.
+// the viewport will scroll to the bottom to show the latest content — EXCEPT
+// when the user is actively selecting text (mouse button held), in which case
+// the scroll position is locked so the highlighted content stays under the
+// cursor. The pending bottom-scroll is deferred to MouseUp.
 func (s *ScrollList) SetItems(items []MessageItem) {
 	s.items = items
-	if s.autoScroll {
+	if s.autoScroll && !s.sel.MouseDown {
 		s.GotoBottom()
 	}
 }
@@ -157,6 +160,10 @@ func (s *ScrollList) HandleMouseDown(x, y int) bool {
 // HandleMouseDrag handles mouse motion while button is held.
 // Updates the selection endpoint for character-level precision.
 // Returns true if selection was updated.
+//
+// Defensively disables auto-scroll on every drag update — even if the
+// MouseDown handler missed (e.g. click landed in viewport padding), any
+// active drag means the user is selecting and the viewport must not jump.
 func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 	if !s.sel.MouseDown {
 		return false
@@ -171,6 +178,9 @@ func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 		return false
 	}
 
+	// Hard-lock the viewport while dragging.
+	s.autoScroll = false
+
 	s.sel.DragItemIdx = itemIdx
 	s.sel.DragLineIdx = lineIdx
 	s.sel.DragCol = x
@@ -178,6 +188,13 @@ func (s *ScrollList) HandleMouseDrag(x, y int) bool {
 	return true
 }
 
+// IsMouseDown reports whether the user currently has the mouse button held
+// (i.e. a selection drag is in progress). Used by the parent model to avoid
+// re-enabling auto-scroll during streaming while the user is selecting.
+func (s *ScrollList) IsMouseDown() bool {
+	return s.sel.MouseDown
+}
+
 // HandleMouseUp handles mouse button release.
 // Returns true if there was an active selection.
 func (s *ScrollList) HandleMouseUp() bool {
@@ -521,6 +538,21 @@ func (s *ScrollList) View() string {
 		for idx := s.offsetIdx; idx < len(s.items) && remainingHeight > 0; idx++ {
 			item := s.items[idx]
 			content := item.Render(s.width)
+
+			// Items that render to an empty string contribute zero height to
+			// the viewport. This MUST match renderedHeight()'s semantics —
+			// otherwise getItemAndLineAtY (which uses renderedHeight) treats
+			// the item as 0 lines while View() emits one blank line via
+			// strings.Split("", "\n") = [""], producing a 1-row downward
+			// drift in mouse hit-testing per empty item between offsetIdx
+			// and the cursor (most visibly streaming-reasoning items before
+			// any reasoning has streamed, which extension widgets surface by
+			// shrinking the scrollback).
+			if content == "" {
+				s.heightCache[item.ID()] = 0
+				continue
+			}
+
 			contentLines := strings.Split(content, "\n")
 
 			// Refresh height cache from the actual render (authoritative).

internal/ui/scrolllist_test.go

@@ -0,0 +1,181 @@
+package ui
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+)
+
+// fakeItem is a deterministic MessageItem for ScrollList tests.
+type fakeItem struct {
+	id    string
+	lines int
+}
+
+func (f *fakeItem) ID() string { return f.id }
+func (f *fakeItem) Render(_ int) string {
+	if f.lines <= 0 {
+		return ""
+	}
+	parts := make([]string, f.lines)
+	for i := range parts {
+		parts[i] = fmt.Sprintf("%s-line-%d", f.id, i)
+	}
+	return strings.Join(parts, "\n")
+}
+func (f *fakeItem) Height() int { return f.lines }
+
+// makeItems builds n fake items of `lines` height each.
+func makeItems(n, lines int) []MessageItem {
+	out := make([]MessageItem, n)
+	for i := range out {
+		out[i] = &fakeItem{id: fmt.Sprintf("item-%d", i), lines: lines}
+	}
+	return out
+}
+
+// TestScrollList_MouseDownPreventsAutoScroll verifies the core fix for the
+// copy-selection drift bug: while the user has the mouse button held
+// (drag-selecting), incoming content updates must NOT shift the viewport,
+// because doing so moves the highlighted content out from under the cursor.
+func TestScrollList_MouseDownPreventsAutoScroll(t *testing.T) {
+	sl := NewScrollList(80, 10)
+	sl.SetItems(makeItems(20, 2)) // 40 lines of content into a 10-line viewport
+	// Capture the auto-scrolled-to-bottom position.
+	startOffsetIdx := sl.offsetIdx
+	startOffsetLine := sl.offsetLine
+
+	// User clicks somewhere in the visible area, starting a drag-select.
+	if !sl.HandleMouseDown(5, 3) {
+		t.Fatalf("HandleMouseDown should accept a click inside the viewport")
+	}
+	if !sl.IsMouseDown() {
+		t.Fatalf("IsMouseDown should be true after HandleMouseDown")
+	}
+
+	// New content arrives. With autoScroll still true, SetItems would
+	// normally call GotoBottom() and shift the viewport. The fix should
+	// suppress that while MouseDown is held.
+	sl.SetItems(makeItems(30, 2)) // 60 lines now
+	if sl.offsetIdx != startOffsetIdx || sl.offsetLine != startOffsetLine {
+		t.Errorf("viewport scrolled during active drag: was (%d,%d), now (%d,%d)",
+			startOffsetIdx, startOffsetLine, sl.offsetIdx, sl.offsetLine)
+	}
+
+	// User releases the mouse — drag is over.
+	sl.HandleMouseUp()
+	if sl.IsMouseDown() {
+		t.Fatalf("IsMouseDown should be false after HandleMouseUp")
+	}
+
+	// After release, a fresh content update should resume auto-scrolling
+	// (move the offset to track the new bottom).
+	afterReleaseIdx := sl.offsetIdx
+	afterReleaseLine := sl.offsetLine
+	sl.SetItems(makeItems(50, 2))
+	if sl.offsetIdx == afterReleaseIdx && sl.offsetLine == afterReleaseLine {
+		t.Errorf("autoscroll did not resume after MouseUp: offset stuck at (%d,%d)",
+			afterReleaseIdx, afterReleaseLine)
+	}
+}
+
+// TestScrollList_DragDisablesAutoScroll verifies that any successful
+// HandleMouseDrag call clears autoScroll, even when HandleMouseDown didn't
+// observe it (e.g. a stale wheel-down event set it back to true mid-stream).
+func TestScrollList_DragDisablesAutoScroll(t *testing.T) {
+	sl := NewScrollList(80, 10)
+	sl.SetItems(makeItems(20, 2))
+
+	// Begin a selection.
+	if !sl.HandleMouseDown(5, 3) {
+		t.Fatalf("HandleMouseDown failed")
+	}
+	// Simulate an external code path that re-enabled autoScroll while
+	// MouseDown is still held (the precise condition that caused drift).
+	sl.autoScroll = true
+
+	// Drag motion should hard-lock the viewport again.
+	if !sl.HandleMouseDrag(10, 4) {
+		t.Fatalf("HandleMouseDrag failed")
+	}
+	if sl.autoScroll {
+		t.Errorf("HandleMouseDrag must clear autoScroll to prevent mid-drag jumps")
+	}
+}
+
+// TestScrollList_SetItemsRespectsMouseDown is the most direct regression
+// test: even with autoScroll enabled and new content appended at the
+// bottom, SetItems must not move the viewport while a mouse drag is in
+// progress. This is what caused the "highlighting shifts by 1+ rows
+// during streaming" symptom reported by the user.
+func TestScrollList_SetItemsRespectsMouseDown(t *testing.T) {
+	sl := NewScrollList(80, 5)
+	sl.SetItems(makeItems(10, 2)) // 20 lines into a 5-line viewport
+	// At bottom.
+	preIdx, preLine := sl.offsetIdx, sl.offsetLine
+
+	// Hold mouse down (no actual drag needed).
+	if !sl.HandleMouseDown(0, 0) {
+		t.Fatalf("HandleMouseDown failed")
+	}
+
+	// Append several more items as if streaming. With the bug, each
+	// SetItems would call GotoBottom and shift the offset.
+	for n := 11; n <= 15; n++ {
+		sl.SetItems(makeItems(n, 2))
+		if sl.offsetIdx != preIdx || sl.offsetLine != preLine {
+			t.Fatalf("viewport drifted during streaming with mouse held: "+
+				"start=(%d,%d) now=(%d,%d) after adding item %d",
+				preIdx, preLine, sl.offsetIdx, sl.offsetLine, n)
+		}
+	}
+}
+
+// TestScrollList_EmptyItemsDoNotShiftMouseMapping is the regression test
+// for the second drift bug: items that render to "" must contribute the
+// same number of rows in View() (zero) as in renderedHeight(), or mouse
+// hit-testing drifts by one row per empty item between offsetIdx and the
+// cursor. This was surfaced by extension widgets (e.g. subagent-monitor)
+// that shrink the scrollback so empty streaming-reasoning items end up
+// in the visible window.
+//
+// Setup: 1 normal item + 1 empty item + 1 normal item. Click on the line
+// where the third item begins. With the bug, getItemAndLineAtY skips the
+// empty item (renderedHeight=0) and reports lineIdx pointing one row
+// past where View() actually painted that line.
+func TestScrollList_EmptyItemsDoNotShiftMouseMapping(t *testing.T) {
+	sl := NewScrollList(80, 10)
+	sl.SetItems([]MessageItem{
+		&fakeItem{id: "a", lines: 2},     // viewY 0–1
+		&fakeItem{id: "empty", lines: 0}, // renders "" — contributes 0 rows
+		&fakeItem{id: "b", lines: 2},     // viewY 2–3
+	})
+
+	// Render the viewport once so the cache reflects what View() actually
+	// emits (this is the path that previously diverged from renderedHeight
+	// for empty items).
+	rendered := sl.View()
+	lines := strings.Split(rendered, "\n")
+
+	// Sanity: View() must emit exactly height lines.
+	if len(lines) != 10 {
+		t.Fatalf("View() returned %d lines, want 10", len(lines))
+	}
+	// Item b's first line should appear at viewY=2, NOT viewY=3.
+	if !strings.Contains(lines[2], "b-line-0") {
+		t.Errorf("viewY=2 should render b-line-0 (empty item contributes 0 rows), got %q", lines[2])
+	}
+
+	// Now the actual hit-test contract: clicking on viewY=2 must map to
+	// item b line 0 — the same coordinate View() rendered there.
+	idx, line := sl.getItemAndLineAtY(2)
+	if idx != 2 || line != 0 {
+		t.Errorf("getItemAndLineAtY(2) = (%d,%d), want (2,0)", idx, line)
+	}
+
+	// And clicking on the second line of b (viewY=3) must map to b line 1.
+	idx, line = sl.getItemAndLineAtY(3)
+	if idx != 2 || line != 1 {
+		t.Errorf("getItemAndLineAtY(3) = (%d,%d), want (2,1)", idx, line)
+	}
+}

internal/ui/selection/selection.go

@@ -230,8 +230,10 @@ func FindWordBoundaries(line string, col int) (startCol, endCol int) {
 
 // HighlightLine applies reverse-video highlighting to a portion of a rendered
 // line (which may contain ANSI escape codes). startCol/endCol are in display
-// columns. If startCol == -1, the entire line is highlighted. If startCol ==
-// endCol, returns the line unchanged.
+// columns. If startCol == -1, the entire line is highlighted. If endCol ==
+// -1, the highlight runs from startCol to the end of the line (the sentinel
+// returned by IsLineInRange for the first line of a multi-line selection).
+// If startCol == endCol, returns the line unchanged.
 //
 // Uses ultraviolet ScreenBuffer for cell-level ANSI manipulation.
 func HighlightLine(line string, startCol, endCol int) string {
@@ -250,6 +252,16 @@ func HighlightLine(line string, startCol, endCol int) string {
 		endCol = lineWidth
 	}
 
+	// "From startCol to end of line" sentinel (returned by IsLineInRange
+	// for the first line of a multi-line selection). Without this branch,
+	// the start line of a multi-line drag would never be highlighted —
+	// the user perceives this as the selection being shifted one row down
+	// from the cursor, especially when extension widgets shrink the
+	// scrollback and make the start line land on a tall styled block.
+	if endCol < 0 {
+		endCol = lineWidth
+	}
+
 	if startCol >= endCol || startCol >= lineWidth {
 		return line
 	}
@@ -296,6 +308,11 @@ func ExtractText(line string, startCol, endCol int) string {
 		endCol = lineWidth
 	}
 
+	// "From startCol to end of line" sentinel (see HighlightLine).
+	if endCol < 0 {
+		endCol = lineWidth
+	}
+
 	if startCol >= endCol || startCol >= lineWidth {
 		return ""
 	}

internal/ui/selection/selection_test.go

@@ -357,6 +357,54 @@ func TestHighlightLine_NoSelection(t *testing.T) {
 	}
 }
 
+// TestHighlightLine_EndOfLineSentinel verifies that endCol=-1 is interpreted
+// as "highlight from startCol to end of line", matching the sentinel
+// returned by IsLineInRange for the first line of a multi-line selection.
+//
+// Regression: without this contract, the start line of any multi-line drag
+// would silently fall through HighlightLine's startCol >= endCol guard and
+// render unstyled, making the selection appear to begin one row below the
+// cursor — the exact "tracking gets shifted" symptom users reported when
+// extension widgets shrank the scrollback enough that the click landed on a
+// styled tool-result block.
+func TestHighlightLine_EndOfLineSentinel(t *testing.T) {
+	line := "Hello, World!"
+	result := HighlightLine(line, 0, -1)
+	if result == line {
+		t.Errorf("endCol=-1 should highlight from startCol to end of line; got unchanged input")
+	}
+	if len(result) <= len(line) {
+		t.Errorf("highlighted result should be longer than plain input (ANSI codes added); got len=%d want > %d", len(result), len(line))
+	}
+}
+
+// TestExtractText_EndOfLineSentinel mirrors TestHighlightLine_EndOfLineSentinel
+// for the extraction path used by the clipboard copy.
+func TestExtractText_EndOfLineSentinel(t *testing.T) {
+	line := "Hello, World!"
+	got := ExtractText(line, 7, -1)
+	want := "World!"
+	if got != want {
+		t.Errorf("ExtractText(line, 7, -1) = %q, want %q", got, want)
+	}
+}
+
+// TestIsLineInRange_StartLineSentinelHighlights composes IsLineInRange with
+// HighlightLine end-to-end: the start line of a multi-line, single-item
+// selection must actually emit highlight ANSI codes. This is the contract
+// the rendering path in scrolllist.View() relies on.
+func TestIsLineInRange_StartLineSentinelHighlights(t *testing.T) {
+	r := Range{StartItemIdx: 5, EndItemIdx: 5, StartLine: 0, EndLine: 2, StartCol: 0, EndCol: 10}
+	inRange, sc, ec := IsLineInRange(r, 5, 0)
+	if !inRange {
+		t.Fatalf("item 5 line 0 should be in range")
+	}
+	highlighted := HighlightLine("first line of selection", sc, ec)
+	if highlighted == "first line of selection" {
+		t.Errorf("first line of multi-line selection was not highlighted (sc=%d ec=%d)", sc, ec)
+	}
+}
+
 // TestMultiClickDetection verifies the click counting logic.
 func TestMultiClickDetection(t *testing.T) {
 	s := NewState()

pkg/kit/README.md

@@ -49,6 +49,36 @@ The SDK behaves identically to the CLI:
 - Respects all environment variables (`KIT_*`)
 - Uses the same defaults as the CLI
 
+Each `kit.New` / `kit.NewAgent` call owns an **isolated configuration store**,
+so constructing multiple Kit instances in the same process is safe — setting
+the model, thinking level, or generation parameters on one never affects
+another, and runtime mutators (`SetModel`, `SetThinkingLevel`) only touch the
+owning instance. This makes subagent spawning and multi-Kit embedding race-free
+without external synchronization.
+
+### Functional options (`NewAgent`)
+
+For simple programmatic setups, `kit.NewAgent` is an ergonomic
+functional-options front door over `kit.New`. Streaming is enabled by default;
+pass `kit.WithStreaming(false)` to opt out.
+
+```go
+host, err := kit.NewAgent(ctx,
+    kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+    kit.WithSystemPrompt("You are a helpful assistant."),
+    kit.WithMaxTokens(8192),
+    kit.WithThinkingLevel("medium"),
+    kit.Ephemeral(), // in-memory session, no persistence
+)
+```
+
+Helpers: `WithModel`, `WithSystemPrompt`, `WithStreaming`, `WithMaxTokens`,
+`WithThinkingLevel`, `WithTools`, `WithExtraTools`, `WithProviderAPIKey`,
+`WithProviderURL`, `WithConfigFile`, `WithDebug`, and `Ephemeral`. `Option` is
+a plain `func(*Options)`, so you can define your own. For fields without a
+`With*` helper (`MCPConfig`, `InProcessMCPServers`, `SessionManager`, MCP task
+tuning) construct an `Options` value and call `kit.New`.
+
 ### Options
 
 You can override specific settings:
@@ -59,7 +89,7 @@ host, err := kit.New(ctx, &kit.Options{
     SystemPrompt: "You are a helpful bot",    // Override system prompt
     ConfigFile:   "/path/to/config.yml",      // Use specific config file
     MaxSteps:     10,                         // Override max steps
-    Streaming:    true,                       // Enable streaming
+    Streaming:    ptrBool(true),               // *bool: nil = unset (default true), &false = off
     Quiet:        true,                       // Suppress debug output
 
     // Session options
@@ -241,9 +271,46 @@ response, _ := host.Prompt(ctx, "What's my name?")
 host.ClearSession()
 ```
 
+### Runtime Skills and Context Files
+
+For multi-tenant chatbots, web services, or any host that needs per-user or
+per-session instructions, the SDK lets you add, remove, and replace skills and
+project context files (e.g. `AGENTS.md`) **after** Kit construction. Every
+mutation recomposes the system prompt and applies it to the agent so the next
+turn picks up the new instructions — no restart required.
+
+```go
+// Add a programmatic skill (no file on disk required).
+host.AddSkill(&kit.Skill{
+    Name:        "polite-french",
+    Description: "Respond in French and always greet the user.",
+    Content:     "Always reply in French. Open every response with 'Bonjour'.",
+})
+
+// Or load one from disk.
+host.LoadAndAddSkill("/var/skills/refund-policy.md")
+
+// Swap per-user AGENTS.md content fetched from your database.
+host.AddContextFileContent(
+    fmt.Sprintf("session://%s/AGENTS.md", userID),
+    rulesFromDB,
+)
+
+// Tear down session-specific state when the user logs off.
+host.RemoveSkill("polite-french")
+host.RemoveContextFile(fmt.Sprintf("session://%s/AGENTS.md", userID))
+
+// Or replace the whole set in one shot.
+host.SetSkills(activeSkillsForUser)
+host.SetContextFiles(activeContextForUser)
+```
+
+Readers (`GetSkills`, `GetContextFiles`) return snapshots, and every mutator
+is safe to call concurrently from multiple goroutines.
+
 ## Re-exported Types
 
-The SDK re-exports types so you don't need direct internal imports:
+The SDK re-exports message/session/MCP types so you don't need direct internal imports. Agent-configuration types are Kit-owned (not aliases) and use only SDK types in their signatures, so consumers never need to import the underlying LLM-provider package.
 
 ```go
 // Message types
@@ -251,13 +318,28 @@ kit.Message, kit.MessageRole, kit.ContentPart
 kit.TextContent, kit.ReasoningContent, kit.ToolCall, kit.ToolResult, kit.Finish
 kit.RoleUser, kit.RoleAssistant, kit.RoleTool, kit.RoleSystem
 
-// LLM types — concrete Kit-owned structs, no external library dependency
+// LLM types — Kit-owned `LLM*` aliases over the underlying provider types,
+// so consumers never import the provider package directly
 kit.LLMMessage      // {Role LLMMessageRole, Content string}
 kit.LLMMessageRole  // "user" | "assistant" | "system" | "tool"
 kit.LLMUsage        // {InputTokens, OutputTokens, TotalTokens, ...}
 kit.LLMResponse     // {Content, FinishReason, Usage}
 kit.LLMFilePart     // {Filename, Data []byte, MediaType}
 
+// Agent configuration — concrete Kit-owned structs and function types.
+// All fields use SDK types (e.g. `[]kit.Tool`), so consumers can construct
+// these without importing any LLM-provider package.
+kit.AgentConfig              // Lower-level agent config — prefer Options unless you need direct control
+kit.DebugLogger              // Interface: LogDebug(string) / IsDebugEnabled() bool
+kit.MCPTaskConfig            // Task-aware MCP tools/call config (modes, polling, progress)
+kit.ToolCallHandler          // func(toolCallID, toolName, toolArgs string)
+kit.ToolExecutionHandler     // func(toolCallID, toolName, toolArgs string, isStarting bool)
+kit.ToolResultHandler        // func(toolCallID, toolName, toolArgs, result, metadata string, isError bool)
+kit.ResponseHandler          // func(content string)
+kit.StreamingResponseHandler // func(content string)
+kit.ToolCallContentHandler   // func(content string)
+kit.SpinnerFunc              // func(fn func() error) error
+
 // MCP OAuth types
 kit.MCPServer            // *server.MCPServer for in-process MCP transport
 kit.MCPServerConfig      // Configuration for an MCP server (stdio, SSE, or in-process)
@@ -279,6 +361,7 @@ msg  := kit.ConvertFromLLMMessage(lMsg)  // LLMMessage  → SDK Message
 
 - `Kit` - Main SDK type
 - `Options` - Configuration options
+- `Option` - Functional option (`func(*Options)`) for `NewAgent`
 - `Message` - Conversation message with typed content parts
 - `Tool` - Agent tool interface
 - `TurnResult` - Full result from a prompt including usage stats
@@ -286,6 +369,7 @@ msg  := kit.ConvertFromLLMMessage(lMsg)  // LLMMessage  → SDK Message
 ### Key Methods
 
 - `New(ctx, opts)` - Create new Kit instance
+- `NewAgent(ctx, ...Option)` - Create a Kit via functional options (streaming on by default)
 - `Prompt(ctx, message)` - Send message and get response string
 - `PromptResult(ctx, message)` - Send message and get full TurnResult
 - `PromptWithOptions(ctx, message, opts)` - Prompt with per-call options
@@ -297,6 +381,9 @@ msg  := kit.ConvertFromLLMMessage(lMsg)  // LLMMessage  → SDK Message
 - `ClearSession()` - Clear conversation history
 - `GetSessionPath()` - Get session file path
 - `GetSessionID()` - Get session UUID
+- `AddSkill(*Skill)` / `LoadAndAddSkill(path)` / `RemoveSkill(name)` / `SetSkills([])` - Manage skills at runtime
+- `AddContextFile(*ContextFile)` / `AddContextFileContent(path, content)` / `LoadAndAddContextFile(path)` / `RemoveContextFile(path)` / `SetContextFiles([])` - Manage AGENTS.md-style context files at runtime
+- `RefreshSystemPrompt()` - Re-apply the composed system prompt to the agent
 - `Close()` - Clean up resources
 
 ### Options

pkg/kit/agent_config_internal_test.go

@@ -0,0 +1,208 @@
+package kit
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/mark3labs/kit/internal/agent"
+)
+
+// TestAgentConfigToInternal verifies that the SDK-side AgentConfig converts
+// faithfully to the internal agent.AgentConfig representation, preserving
+// every field consumed by the internal agent layer.
+//
+// Regression test for https://github.com/mark3labs/kit/issues/30.
+func TestAgentConfigToInternal(t *testing.T) {
+	t.Run("nil receiver returns nil", func(t *testing.T) {
+		var c *AgentConfig
+		if got := c.toInternal(); got != nil {
+			t.Errorf("nil.toInternal() = %v, want nil", got)
+		}
+	})
+
+	t.Run("scalar fields round-trip", func(t *testing.T) {
+		c := &AgentConfig{
+			SystemPrompt:     "sys",
+			MaxSteps:         7,
+			StreamingEnabled: true,
+			DisableCoreTools: true,
+		}
+		got := c.toInternal()
+		if got == nil {
+			t.Fatal("toInternal() = nil")
+		}
+		if got.SystemPrompt != "sys" {
+			t.Errorf("SystemPrompt = %q, want %q", got.SystemPrompt, "sys")
+		}
+		if got.MaxSteps != 7 {
+			t.Errorf("MaxSteps = %d, want 7", got.MaxSteps)
+		}
+		if !got.StreamingEnabled {
+			t.Error("StreamingEnabled = false, want true")
+		}
+		if !got.DisableCoreTools {
+			t.Error("DisableCoreTools = false, want true")
+		}
+	})
+
+	t.Run("tool slices propagate without conversion", func(t *testing.T) {
+		// Tool is a type alias for the underlying LLM-tool type, so the
+		// SDK []Tool and internal []fantasy.AgentTool slices share the
+		// same backing array after conversion.
+		tool := NewTool[struct{}]("noop", "noop", nil)
+		c := &AgentConfig{
+			CoreTools:  []Tool{tool},
+			ExtraTools: []Tool{tool, tool},
+		}
+		got := c.toInternal()
+		if len(got.CoreTools) != 1 {
+			t.Errorf("CoreTools len = %d, want 1", len(got.CoreTools))
+		}
+		if len(got.ExtraTools) != 2 {
+			t.Errorf("ExtraTools len = %d, want 2", len(got.ExtraTools))
+		}
+	})
+
+	t.Run("tool wrapper is invoked through internal config", func(t *testing.T) {
+		called := false
+		c := &AgentConfig{
+			ToolWrapper: func(in []Tool) []Tool {
+				called = true
+				return in
+			},
+		}
+		got := c.toInternal()
+		if got.ToolWrapper == nil {
+			t.Fatal("internal ToolWrapper is nil")
+		}
+		_ = got.ToolWrapper(nil)
+		if !called {
+			t.Error("SDK ToolWrapper was not invoked through the internal config")
+		}
+	})
+
+	t.Run("OnMCPServerLoaded propagates", func(t *testing.T) {
+		var captured string
+		wantErr := errors.New("boom")
+		c := &AgentConfig{
+			OnMCPServerLoaded: func(name string, _ int, _ error) {
+				captured = name
+			},
+		}
+		got := c.toInternal()
+		got.OnMCPServerLoaded("svr", 3, wantErr)
+		if captured != "svr" {
+			t.Errorf("OnMCPServerLoaded captured = %q, want %q", captured, "svr")
+		}
+	})
+
+	t.Run("DebugLogger propagates", func(t *testing.T) {
+		dl := &fakeDebugLogger{enabled: true}
+		c := &AgentConfig{DebugLogger: dl}
+		got := c.toInternal()
+		if got.DebugLogger == nil {
+			t.Fatal("internal DebugLogger is nil")
+		}
+		if !got.DebugLogger.IsDebugEnabled() {
+			t.Error("IsDebugEnabled = false, want true")
+		}
+		got.DebugLogger.LogDebug("hello")
+		if len(dl.messages) != 1 || dl.messages[0] != "hello" {
+			t.Errorf("messages = %v, want [hello]", dl.messages)
+		}
+	})
+
+	t.Run("MCPTaskConfig propagates with mode + progress", func(t *testing.T) {
+		c := &AgentConfig{
+			MCPTaskConfig: MCPTaskConfig{
+				PerServerMode: map[string]MCPTaskMode{
+					"build-svr": MCPTaskModeAlways,
+				},
+				DefaultTTL:      30 * time.Second,
+				PollInterval:    250 * time.Millisecond,
+				MaxPollInterval: 2 * time.Second,
+				Timeout:         5 * time.Minute,
+				Progress:        func(_ MCPTaskProgress) {},
+			},
+		}
+		got := c.toInternal()
+		if got.MCPTaskConfig.DefaultTTL != 30*time.Second {
+			t.Errorf("DefaultTTL = %v, want 30s", got.MCPTaskConfig.DefaultTTL)
+		}
+		if got.MCPTaskConfig.PollInterval != 250*time.Millisecond {
+			t.Errorf("PollInterval = %v, want 250ms", got.MCPTaskConfig.PollInterval)
+		}
+		if got.MCPTaskConfig.MaxPollInterval != 2*time.Second {
+			t.Errorf("MaxPollInterval = %v, want 2s", got.MCPTaskConfig.MaxPollInterval)
+		}
+		if got.MCPTaskConfig.Timeout != 5*time.Minute {
+			t.Errorf("Timeout = %v, want 5m", got.MCPTaskConfig.Timeout)
+		}
+		mode, ok := got.MCPTaskConfig.PerServerMode["build-svr"]
+		if !ok {
+			t.Fatal("PerServerMode missing 'build-svr'")
+		}
+		if string(mode) != string(MCPTaskModeAlways) {
+			t.Errorf("mode = %q, want %q", mode, MCPTaskModeAlways)
+		}
+		if got.MCPTaskConfig.Progress == nil {
+			t.Fatal("internal Progress handler is nil")
+		}
+	})
+
+	t.Run("auth and token store factories are wired", func(t *testing.T) {
+		auth := &fakeAuthHandler{}
+		tokenCalls := 0
+		var tokenServer string
+		factory := MCPTokenStoreFactory(func(server string) (MCPTokenStore, error) {
+			tokenCalls++
+			tokenServer = server
+			return nil, nil
+		})
+		c := &AgentConfig{
+			AuthHandler:       auth,
+			TokenStoreFactory: factory,
+		}
+		got := c.toInternal()
+		if got.AuthHandler == nil {
+			t.Fatal("internal AuthHandler is nil")
+		}
+		if got.TokenStoreFactory == nil {
+			t.Fatal("internal TokenStoreFactory is nil")
+		}
+		_, _ = got.TokenStoreFactory("https://example.test")
+		if tokenCalls != 1 {
+			t.Errorf("token factory call count = %d, want 1", tokenCalls)
+		}
+		if tokenServer != "https://example.test" {
+			t.Errorf("token factory server arg = %q", tokenServer)
+		}
+		if got.AuthHandler.RedirectURI() != "redirect" {
+			t.Errorf("RedirectURI = %q, want %q", got.AuthHandler.RedirectURI(), "redirect")
+		}
+	})
+
+	// Compile-time check that the internal type is what we expect.
+	//nolint:staticcheck // QF1011: explicit type asserts the conversion target.
+	var _ *agent.AgentConfig = (&AgentConfig{}).toInternal()
+}
+
+// fakeAuthHandler implements both kit.MCPAuthHandler and the structurally
+// identical tools.MCPAuthHandler used by the internal layer.
+type fakeAuthHandler struct{}
+
+func (f *fakeAuthHandler) RedirectURI() string { return "redirect" }
+func (f *fakeAuthHandler) HandleAuth(_ context.Context, _ string, _ string) (string, error) {
+	return "", nil
+}
+
+// fakeDebugLogger implements kit.DebugLogger for tests.
+type fakeDebugLogger struct {
+	enabled  bool
+	messages []string
+}
+
+func (f *fakeDebugLogger) LogDebug(m string)    { f.messages = append(f.messages, m) }
+func (f *fakeDebugLogger) IsDebugEnabled() bool { return f.enabled }

pkg/kit/config.go

@@ -65,23 +65,46 @@ const sdkDefaultMaxTokens = 8192
 //     which returns models.ThinkingOff.
 //   - sampling params (temperature, top-p, top-k, frequency/presence-penalty):
 //     left as nil pointers so provider libraries apply their own defaults.
-func setSDKDefaults() {
-	viper.SetDefault("model", "anthropic/claude-sonnet-4-5-20250929")
-	viper.SetDefault("system-prompt", defaultSystemPrompt)
-	viper.SetDefault("stream", true)
-	viper.SetDefault("num-gpu-layers", -1)
-	viper.SetDefault("main-gpu", 0)
+func setSDKDefaults(v *viper.Viper) {
+	v.SetDefault("model", "anthropic/claude-sonnet-4-5-20250929")
+	v.SetDefault("system-prompt", defaultSystemPrompt)
+	v.SetDefault("stream", true)
+	v.SetDefault("num-gpu-layers", -1)
+	v.SetDefault("main-gpu", 0)
 }
 
-// InitConfig initializes the viper configuration system.
+// InitConfig initializes the process-global viper configuration system.
 // It searches for config files in standard locations and loads them with
 // environment variable substitution.
 //
 // configFile: explicit config file path (empty = search defaults).
 // debug: if true, print warnings about missing configs to stderr.
+//
+// This wraps [initConfig] using the process-global store and is retained for
+// the CLI, which binds its flags to the global viper.
 func InitConfig(configFile string, debug bool) error {
+	return initConfig(viper.GetViper(), configFile, debug)
+}
+
+// initConfig loads configuration into the supplied per-instance store. When v
+// is nil the process-global store is used.
+func initConfig(v *viper.Viper, configFile string, debug bool) error {
+	if v == nil {
+		v = viper.GetViper()
+	}
+
+	// Configure KIT_* environment overrides unconditionally, before any file
+	// is loaded, so that an explicit config file does not disable env support.
+	// Map hyphenated config keys (e.g. "max-tokens") to underscored env var
+	// names (e.g. KIT_MAX_TOKENS); without this AutomaticEnv looks for
+	// KIT_MAX-TOKENS and silently misses valid overrides. Precedence is
+	// resolved at read time, so calling these before ReadConfig is fine.
+	v.SetEnvPrefix("KIT")
+	v.SetEnvKeyReplacer(strings.NewReplacer("-", "_"))
+	v.AutomaticEnv()
+
 	if configFile != "" {
-		return LoadConfigWithEnvSubstitution(configFile)
+		return loadConfigWithEnvSubstitution(v, configFile)
 	}
 
 	// Ensure a config file exists (create default if none found).
@@ -97,15 +120,15 @@ func InitConfig(configFile string, debug bool) error {
 	}
 
 	// Current directory has higher priority than home directory.
-	viper.AddConfigPath(".")
-	viper.AddConfigPath(home)
+	v.AddConfigPath(".")
+	v.AddConfigPath(home)
 
 	configLoaded := false
 
-	viper.SetConfigName(".kit")
-	if err := viper.ReadInConfig(); err == nil {
-		configPath := viper.ConfigFileUsed()
-		if err := LoadConfigWithEnvSubstitution(configPath); err != nil {
+	v.SetConfigName(".kit")
+	if err := v.ReadInConfig(); err == nil {
+		configPath := v.ConfigFileUsed()
+		if err := loadConfigWithEnvSubstitution(v, configPath); err != nil {
 			if strings.Contains(err.Error(), "environment variable substitution failed") {
 				return fmt.Errorf("error reading config file '%s': %w", configPath, err)
 			}
@@ -118,17 +141,21 @@ func InitConfig(configFile string, debug bool) error {
 		fmt.Fprintf(os.Stderr, "No config file found in current directory or home directory\n")
 	}
 
-	viper.SetEnvPrefix("KIT")
-	// Map hyphenated config keys (e.g. "max-tokens") to underscored env
-	// var names (e.g. KIT_MAX_TOKENS). Without this, AutomaticEnv looks
-	// for KIT_MAX-TOKENS and silently misses valid env overrides.
-	viper.SetEnvKeyReplacer(strings.NewReplacer("-", "_"))
-	viper.AutomaticEnv()
 	return nil
 }
 
-// LoadConfigWithEnvSubstitution loads a config file with ${ENV_VAR} expansion.
+// LoadConfigWithEnvSubstitution loads a config file with ${ENV_VAR} expansion
+// into the process-global viper store.
 func LoadConfigWithEnvSubstitution(configPath string) error {
+	return loadConfigWithEnvSubstitution(viper.GetViper(), configPath)
+}
+
+// loadConfigWithEnvSubstitution loads a config file with ${ENV_VAR} expansion
+// into the supplied per-instance store (or the global store when v is nil).
+func loadConfigWithEnvSubstitution(v *viper.Viper, configPath string) error {
+	if v == nil {
+		v = viper.GetViper()
+	}
 	rawContent, err := os.ReadFile(configPath)
 	if err != nil {
 		return fmt.Errorf("failed to read config file: %w", err)
@@ -146,6 +173,6 @@ func LoadConfigWithEnvSubstitution(configPath string) error {
 	}
 
 	config.SetConfigPath(configPath)
-	viper.SetConfigType(configType)
-	return viper.ReadConfig(strings.NewReader(processedContent))
+	v.SetConfigType(configType)
+	return v.ReadConfig(strings.NewReader(processedContent))
 }

pkg/kit/context_files.go

@@ -0,0 +1,150 @@
+package kit
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+)
+
+// ---------------------------------------------------------------------------
+// Runtime context-file management (Issue #36)
+// ---------------------------------------------------------------------------
+//
+// Project context files (AGENTS.md and friends) are normally auto-discovered
+// during Kit.New() and injected into the system prompt. SDK consumers building
+// multi-tenant chatbots often need to swap context per user/session at runtime
+// without restarting the agent. The methods below provide that surface.
+//
+// Every mutation recomposes the system prompt and applies it to the underlying
+// agent so the next turn sees the updated project context.
+
+// AddContextFile registers a project context file (e.g. an AGENTS.md
+// equivalent) on this Kit instance. The file does not need to exist on
+// disk — Path is treated as an opaque identifier used both for de-duplication
+// and for the "Instructions from: <Path>" header injected into the system
+// prompt. If a context file with the same Path is already loaded the new
+// content replaces it.
+//
+// Returns an error when cf is nil or has an empty Path. AddContextFile is
+// safe to call from any goroutine.
+func (m *Kit) AddContextFile(cf *ContextFile) error {
+	if cf == nil {
+		return fmt.Errorf("AddContextFile: context file is nil")
+	}
+	if cf.Path == "" {
+		return fmt.Errorf("AddContextFile: context file path is required")
+	}
+
+	// Take a defensive copy so later mutations by the caller don't race with
+	// the agent reading the composed prompt.
+	stored := &ContextFile{
+		Path:    cf.Path,
+		Content: strings.TrimSpace(cf.Content),
+	}
+
+	m.runtimeMu.Lock()
+	replaced := false
+	for i, existing := range m.contextFiles {
+		if existing.Path == stored.Path {
+			m.contextFiles[i] = stored
+			replaced = true
+			break
+		}
+	}
+	if !replaced {
+		m.contextFiles = append(m.contextFiles, stored)
+	}
+	m.runtimeMu.Unlock()
+
+	m.applyComposedSystemPrompt()
+	return nil
+}
+
+// AddContextFileContent is a convenience wrapper around [Kit.AddContextFile]
+// that builds the ContextFile from a path and inline content string. Use this
+// when the context originates from a database, API response, or any other
+// non-filesystem source.
+func (m *Kit) AddContextFileContent(path, content string) (*ContextFile, error) {
+	cf := &ContextFile{Path: path, Content: content}
+	if err := m.AddContextFile(cf); err != nil {
+		return nil, err
+	}
+	return cf, nil
+}
+
+// LoadAndAddContextFile reads a file from disk and registers it as a project
+// context file via [Kit.AddContextFile]. The absolute path is stored on the
+// resulting ContextFile.
+func (m *Kit) LoadAndAddContextFile(path string) (*ContextFile, error) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return nil, fmt.Errorf("LoadAndAddContextFile: %w", err)
+	}
+	abs, absErr := filepath.Abs(path)
+	if absErr != nil {
+		abs = path
+	}
+	cf := &ContextFile{
+		Path:    abs,
+		Content: strings.TrimSpace(string(data)),
+	}
+	if err := m.AddContextFile(cf); err != nil {
+		return nil, err
+	}
+	return cf, nil
+}
+
+// RemoveContextFile removes the context file with the given path and
+// recomposes the system prompt. Returns true when a matching file was found
+// and removed, false otherwise.
+func (m *Kit) RemoveContextFile(path string) bool {
+	m.runtimeMu.Lock()
+	found := false
+	for i, cf := range m.contextFiles {
+		if cf.Path == path {
+			m.contextFiles = append(m.contextFiles[:i], m.contextFiles[i+1:]...)
+			found = true
+			break
+		}
+	}
+	m.runtimeMu.Unlock()
+
+	if !found {
+		return false
+	}
+	m.applyComposedSystemPrompt()
+	return true
+}
+
+// SetContextFiles replaces the active context-file set with the provided
+// slice. Pass nil or an empty slice to clear all context. The system prompt
+// is recomposed and applied. ContextFiles with empty Paths are rejected and
+// no mutation is performed.
+func (m *Kit) SetContextFiles(files []*ContextFile) error {
+	// Validate first so a bad input doesn't partially mutate state.
+	for i, cf := range files {
+		if cf == nil {
+			return fmt.Errorf("SetContextFiles: context file at index %d is nil", i)
+		}
+		if cf.Path == "" {
+			return fmt.Errorf("SetContextFiles: context file at index %d has empty path", i)
+		}
+	}
+
+	// Defensive copies so caller-side mutation cannot race with composition.
+	copied := make([]*ContextFile, len(files))
+	for i, cf := range files {
+		copied[i] = &ContextFile{
+			Path:    cf.Path,
+			Content: strings.TrimSpace(cf.Content),
+		}
+	}
+
+	m.runtimeMu.Lock()
+	m.contextFiles = copied
+	m.runtimeMu.Unlock()
+
+	m.applyComposedSystemPrompt()
+	return nil
+}

pkg/kit/export_test.go

@@ -0,0 +1,22 @@
+package kit
+
+// This file exposes a handful of internal accessors to the external kit_test
+// package. Because it ends in _test.go it is only compiled during testing and
+// is therefore not part of the public SDK surface.
+
+// ConfigValueIsSetForTest reports whether key is explicitly set in this Kit's
+// isolated configuration store. Used by tests to assert the tri-state
+// precedence contract per-instance.
+func (m *Kit) ConfigValueIsSetForTest(key string) bool { return m.v.IsSet(key) }
+
+// ConfigStringForTest returns the string value of key from this Kit's isolated
+// configuration store.
+func (m *Kit) ConfigStringForTest(key string) string { return m.v.GetString(key) }
+
+// ConfigFloatForTest returns the float64 value of key from this Kit's isolated
+// configuration store.
+func (m *Kit) ConfigFloatForTest(key string) float64 { return m.v.GetFloat64(key) }
+
+// ConfigBoolForTest returns the bool value of key from this Kit's isolated
+// configuration store.
+func (m *Kit) ConfigBoolForTest(key string) bool { return m.v.GetBool(key) }

pkg/kit/extension_api.go

@@ -8,55 +8,104 @@ import (
 	"github.com/mark3labs/kit/internal/session"
 )
 
+// ==== Extension Types ====
+//
+// Type aliases for internal extension types exposed through the public
+// ExtensionAPI interface. External SDK consumers can use these without
+// importing internal packages directly.
+
+// ExtensionContext holds the runtime context passed to extensions, including
+// callbacks for printing, sending messages, and accessing session state.
+type ExtensionContext = extensions.Context
+
+// ExtensionWidgetConfig describes a widget registered by an extension.
+type ExtensionWidgetConfig = extensions.WidgetConfig
+
+// ExtensionWidgetPlacement indicates where a widget should be rendered
+// (e.g. above or below the conversation).
+type ExtensionWidgetPlacement = extensions.WidgetPlacement
+
+// ExtensionHeaderFooterConfig describes a header or footer registered by an extension.
+type ExtensionHeaderFooterConfig = extensions.HeaderFooterConfig
+
+// ExtensionEditorConfig configures editor behaviour overrides set by extensions.
+type ExtensionEditorConfig = extensions.EditorConfig
+
+// ExtensionUIVisibility controls which UI elements are visible.
+type ExtensionUIVisibility = extensions.UIVisibility
+
+// ExtensionToolRenderConfig describes custom tool output rendering registered by an extension.
+type ExtensionToolRenderConfig = extensions.ToolRenderConfig
+
+// ExtensionMessageRendererConfig describes custom message rendering registered by an extension.
+type ExtensionMessageRendererConfig = extensions.MessageRendererConfig
+
+// ExtensionSessionMessage represents a single message in the session history
+// as exposed to extensions.
+type ExtensionSessionMessage = extensions.SessionMessage
+
+// ExtensionEntry represents a custom data entry stored by an extension
+// in the session tree.
+type ExtensionEntry = extensions.ExtensionEntry
+
+// ExtensionStatusBarEntry describes a status bar entry registered by an extension.
+type ExtensionStatusBarEntry = extensions.StatusBarEntry
+
+// ExtensionToolInfo describes a tool available to the agent, as seen by extensions.
+type ExtensionToolInfo = extensions.ToolInfo
+
+// ExtensionCommandDef describes a slash command registered by an extension.
+type ExtensionCommandDef = extensions.CommandDef
+
 // ExtensionAPI provides grouped access to all extension-related functionality.
 // This cleans up the main Kit API surface while keeping all extension capabilities available.
 type ExtensionAPI interface {
 	// Context management
-	SetContext(ctx extensions.Context)
-	GetContext() extensions.Context
+	SetContext(ctx ExtensionContext)
+	GetContext() ExtensionContext
 	UpdateContextModel(model string)
 
 	// Widgets
-	SetWidget(config extensions.WidgetConfig)
+	SetWidget(config ExtensionWidgetConfig)
 	RemoveWidget(id string)
-	GetWidgets(placement extensions.WidgetPlacement) []extensions.WidgetConfig
+	GetWidgets(placement ExtensionWidgetPlacement) []ExtensionWidgetConfig
 
 	// Header/Footer
-	SetHeader(config extensions.HeaderFooterConfig)
+	SetHeader(config ExtensionHeaderFooterConfig)
 	RemoveHeader()
-	GetHeader() *extensions.HeaderFooterConfig
-	SetFooter(config extensions.HeaderFooterConfig)
+	GetHeader() *ExtensionHeaderFooterConfig
+	SetFooter(config ExtensionHeaderFooterConfig)
 	RemoveFooter()
-	GetFooter() *extensions.HeaderFooterConfig
+	GetFooter() *ExtensionHeaderFooterConfig
 
 	// Editor
-	SetEditor(config extensions.EditorConfig)
+	SetEditor(config ExtensionEditorConfig)
 	ResetEditor()
-	GetEditor() *extensions.EditorConfig
+	GetEditor() *ExtensionEditorConfig
 
 	// UI Visibility
-	SetUIVisibility(v extensions.UIVisibility)
-	GetUIVisibility() *extensions.UIVisibility
+	SetUIVisibility(v ExtensionUIVisibility)
+	GetUIVisibility() *ExtensionUIVisibility
 
 	// Tool rendering
-	GetToolRenderer(toolName string) *extensions.ToolRenderConfig
-	GetMessageRenderer(name string) *extensions.MessageRendererConfig
+	GetToolRenderer(toolName string) *ExtensionToolRenderConfig
+	GetMessageRenderer(name string) *ExtensionMessageRendererConfig
 
 	// Session data
-	GetSessionMessages() []extensions.SessionMessage
+	GetSessionMessages() []ExtensionSessionMessage
 	AppendEntry(extType, data string) (string, error)
-	GetEntries(extType string) []extensions.ExtensionEntry
+	GetEntries(extType string) []ExtensionEntry
 
 	// Status bar
-	SetStatus(entry extensions.StatusBarEntry)
+	SetStatus(entry ExtensionStatusBarEntry)
 	RemoveStatus(key string)
-	GetStatusEntries() []extensions.StatusBarEntry
+	GetStatusEntries() []ExtensionStatusBarEntry
 
 	// Shortcuts
 	GetShortcuts() map[string]func()
 
 	// Tools
-	GetToolInfos() []extensions.ToolInfo
+	GetToolInfos() []ExtensionToolInfo
 	SetActiveTools(names []string)
 
 	// Options
@@ -71,11 +120,27 @@ type ExtensionAPI interface {
 	EmitBeforeSessionSwitch(switchReason string) (cancelled bool, reason string)
 
 	// Commands
-	Commands() []extensions.CommandDef
+	Commands() []ExtensionCommandDef
 
 	// Lifecycle
 	Reload() error
 	HasExtensions() bool
+
+	// Loaded returns metadata about the extensions currently loaded.
+	Loaded() []ExtensionInfo
+}
+
+// ExtensionInfo describes a single loaded extension for display purposes
+// (e.g. the startup banner or `kit extensions list`).
+type ExtensionInfo struct {
+	// Path is the absolute path of the extension's .go file.
+	Path string
+	// ToolCount is the number of tools registered by the extension.
+	ToolCount int
+	// CommandCount is the number of slash commands registered.
+	CommandCount int
+	// HandlerCount is the total number of event handlers registered.
+	HandlerCount int
 }
 
 // extensionAPI implements ExtensionAPI by wrapping a Kit instance.
@@ -456,3 +521,27 @@ func (e *extensionAPI) Reload() error {
 func (e *extensionAPI) HasExtensions() bool {
 	return e.kit.extRunner != nil
 }
+
+func (e *extensionAPI) Loaded() []ExtensionInfo {
+	if e.kit.extRunner == nil {
+		return nil
+	}
+	exts := e.kit.extRunner.Extensions()
+	if len(exts) == 0 {
+		return nil
+	}
+	infos := make([]ExtensionInfo, 0, len(exts))
+	for _, ex := range exts {
+		handlerCount := 0
+		for _, hs := range ex.Handlers {
+			handlerCount += len(hs)
+		}
+		infos = append(infos, ExtensionInfo{
+			Path:         ex.Path,
+			ToolCount:    len(ex.Tools),
+			CommandCount: len(ex.Commands),
+			HandlerCount: handlerCount,
+		})
+	}
+	return infos
+}

pkg/kit/extensions_bridge.go

@@ -54,83 +54,51 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 	// Subscribe to SDK events and forward to extension runner so extensions
 	// see lifecycle events from the SDK's runTurn()/generate() path.
 
-	if runner.HasHandlers(extensions.AgentStart) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(TurnStartEvent); ok {
-				_, _ = runner.Emit(extensions.AgentStartEvent{Prompt: ev.Prompt})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.AgentStart, func(ev TurnStartEvent) extensions.Event {
+		return extensions.AgentStartEvent{Prompt: ev.Prompt}
+	})
 
-	if runner.HasHandlers(extensions.MessageStart) {
-		m.Subscribe(func(e Event) {
-			if _, ok := e.(MessageStartEvent); ok {
-				_, _ = runner.Emit(extensions.MessageStartEvent{})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.MessageStart, func(_ MessageStartEvent) extensions.Event {
+		return extensions.MessageStartEvent{}
+	})
 
-	if runner.HasHandlers(extensions.MessageUpdate) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(MessageUpdateEvent); ok {
-				_, _ = runner.Emit(extensions.MessageUpdateEvent{Chunk: ev.Chunk})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.MessageUpdate, func(ev MessageUpdateEvent) extensions.Event {
+		return extensions.MessageUpdateEvent{Chunk: ev.Chunk}
+	})
 
-	if runner.HasHandlers(extensions.MessageEnd) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(MessageEndEvent); ok {
-				_, _ = runner.Emit(extensions.MessageEndEvent{Content: ev.Content})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.MessageEnd, func(ev MessageEndEvent) extensions.Event {
+		return extensions.MessageEndEvent{Content: ev.Content}
+	})
 
 	// Tool output streaming events (observation only).
-	if runner.HasHandlers(extensions.ToolOutput) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ToolOutputEvent); ok {
-				_, _ = runner.Emit(extensions.ToolOutputEvent{
-					ToolCallID: ev.ToolCallID,
-					ToolName:   ev.ToolName,
-					Chunk:      ev.Chunk,
-					IsStderr:   ev.IsStderr,
-				})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.ToolOutput, func(ev ToolOutputEvent) extensions.Event {
+		return extensions.ToolOutputEvent{
+			ToolCallID: ev.ToolCallID,
+			ToolName:   ev.ToolName,
+			Chunk:      ev.Chunk,
+			IsStderr:   ev.IsStderr,
+		}
+	})
 
 	// Tool call input streaming events — fire as the LLM generates tool arguments.
-	if runner.HasHandlers(extensions.ToolCallInputStart) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ToolCallStartEvent); ok {
-				_, _ = runner.Emit(extensions.ToolCallInputStartEvent{
-					ToolCallID: ev.ToolCallID,
-					ToolName:   ev.ToolName,
-					ToolKind:   ev.ToolKind,
-				})
-			}
-		})
-	}
-	if runner.HasHandlers(extensions.ToolCallInputDelta) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ToolCallDeltaEvent); ok {
-				_, _ = runner.Emit(extensions.ToolCallInputDeltaEvent{
-					ToolCallID: ev.ToolCallID,
-					Delta:      ev.Delta,
-				})
-			}
-		})
-	}
-	if runner.HasHandlers(extensions.ToolCallInputEnd) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ToolCallEndEvent); ok {
-				_, _ = runner.Emit(extensions.ToolCallInputEndEvent{
-					ToolCallID: ev.ToolCallID,
-				})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.ToolCallInputStart, func(ev ToolCallStartEvent) extensions.Event {
+		return extensions.ToolCallInputStartEvent{
+			ToolCallID: ev.ToolCallID,
+			ToolName:   ev.ToolName,
+			ToolKind:   ev.ToolKind,
+		}
+	})
+	bridgeObserve(m, runner, extensions.ToolCallInputDelta, func(ev ToolCallDeltaEvent) extensions.Event {
+		return extensions.ToolCallInputDeltaEvent{
+			ToolCallID: ev.ToolCallID,
+			Delta:      ev.Delta,
+		}
+	})
+	bridgeObserve(m, runner, extensions.ToolCallInputEnd, func(ev ToolCallEndEvent) extensions.Event {
+		return extensions.ToolCallInputEndEvent{
+			ToolCallID: ev.ToolCallID,
+		}
+	})
 
 	if runner.HasHandlers(extensions.AgentEnd) {
 		m.Subscribe(func(e Event) {
@@ -278,54 +246,13 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 	// Extension ContextPrepare → SDK ContextPrepare hook.
 	if runner.HasHandlers(extensions.ContextPrepare) {
 		m.OnContextPrepare(HookPriorityNormal, func(h ContextPrepareHook) *ContextPrepareResult {
-			// Convert LLM message slice to extension ContextMessage slice.
-			// Extract plain text from each message for the extension API.
-			extMsgs := make([]extensions.ContextMessage, len(h.Messages))
-			for i, msg := range h.Messages {
-				var sb strings.Builder
-				for _, part := range msg.Content {
-					if tp, ok := part.(LLMTextPart); ok {
-						sb.WriteString(tp.Text)
-					}
-				}
-				extMsgs[i] = extensions.ContextMessage{
-					Index:   i,
-					Role:    string(msg.Role),
-					Content: sb.String(),
-				}
-			}
-
+			extMsgs := llmToContextMessages(h.Messages)
 			result, _ := runner.Emit(extensions.ContextPrepareEvent{Messages: extMsgs})
 			r, ok := result.(extensions.ContextPrepareResult)
 			if !ok || r.Messages == nil {
 				return nil
 			}
-
-			// Rebuild LLM message slice from extension result.
-			rebuilt := make([]LLMMessage, 0, len(r.Messages))
-			for _, cm := range r.Messages {
-				if cm.Index >= 0 && cm.Index < len(h.Messages) {
-					// Reuse original message (preserves original role and content).
-					rebuilt = append(rebuilt, h.Messages[cm.Index])
-				} else {
-					// New message injected by extension — construct from role + text.
-					role := LLMRoleUser
-					switch cm.Role {
-					case "assistant":
-						role = LLMRoleAssistant
-					case "system":
-						role = LLMRoleSystem
-					case "tool":
-						role = LLMRoleTool
-					}
-					rebuilt = append(rebuilt, LLMMessage{
-						Role:    role,
-						Content: []LLMMessagePart{LLMTextPart{Text: cm.Content}},
-					})
-				}
-			}
-
-			return &ContextPrepareResult{Messages: rebuilt}
+			return &ContextPrepareResult{Messages: contextMessagesToLLM(r.Messages, h.Messages)}
 		})
 	}
 
@@ -359,99 +286,56 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 
 	// --- Step lifecycle observation events ---
 
-	if runner.HasHandlers(extensions.StepStart) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(StepStartEvent); ok {
-				_, _ = runner.Emit(extensions.StepStartEvent{StepNumber: ev.StepNumber})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.StepStart, func(ev StepStartEvent) extensions.Event {
+		return extensions.StepStartEvent{StepNumber: ev.StepNumber}
+	})
 
-	if runner.HasHandlers(extensions.StepFinish) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(StepFinishEvent); ok {
-				_, _ = runner.Emit(extensions.StepFinishEvent{
-					StepNumber:       ev.StepNumber,
-					HasToolCalls:     ev.HasToolCalls,
-					FinishReason:     ev.FinishReason,
-					InputTokens:      ev.Usage.InputTokens,
-					OutputTokens:     ev.Usage.OutputTokens,
-					CacheReadTokens:  ev.Usage.CacheReadTokens,
-					CacheWriteTokens: ev.Usage.CacheCreationTokens,
-				})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.StepFinish, func(ev StepFinishEvent) extensions.Event {
+		return extensions.StepFinishEvent{
+			StepNumber:       ev.StepNumber,
+			HasToolCalls:     ev.HasToolCalls,
+			FinishReason:     ev.FinishReason,
+			InputTokens:      ev.Usage.InputTokens,
+			OutputTokens:     ev.Usage.OutputTokens,
+			CacheReadTokens:  ev.Usage.CacheReadTokens,
+			CacheWriteTokens: ev.Usage.CacheCreationTokens,
+		}
+	})
 
-	if runner.HasHandlers(extensions.ReasoningStart) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ReasoningStartEvent); ok {
-				_, _ = runner.Emit(extensions.ReasoningStartEvent{ID: ev.ID})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.ReasoningStart, func(ev ReasoningStartEvent) extensions.Event {
+		return extensions.ReasoningStartEvent{ID: ev.ID}
+	})
 
-	if runner.HasHandlers(extensions.Warnings) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(WarningsEvent); ok {
-				_, _ = runner.Emit(extensions.WarningsEvent{Warnings: ev.Warnings})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.Warnings, func(ev WarningsEvent) extensions.Event {
+		return extensions.WarningsEvent{Warnings: ev.Warnings}
+	})
 
-	if runner.HasHandlers(extensions.Source) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(SourceEvent); ok {
-				_, _ = runner.Emit(extensions.SourceEvent{
-					SourceType: ev.SourceType,
-					ID:         ev.ID,
-					URL:        ev.URL,
-					Title:      ev.Title,
-				})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.Source, func(ev SourceEvent) extensions.Event {
+		return extensions.SourceEvent{
+			SourceType: ev.SourceType,
+			ID:         ev.ID,
+			URL:        ev.URL,
+			Title:      ev.Title,
+		}
+	})
 
-	if runner.HasHandlers(extensions.Error) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(ErrorEvent); ok {
-				_, _ = runner.Emit(extensions.ErrorEvent{Error: ev.Error.Error()})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.Error, func(ev ErrorEvent) extensions.Event {
+		return extensions.ErrorEvent{Error: ev.Error.Error()}
+	})
 
-	if runner.HasHandlers(extensions.Retry) {
-		m.Subscribe(func(e Event) {
-			if ev, ok := e.(RetryEvent); ok {
-				_, _ = runner.Emit(extensions.RetryEvent{
-					Attempt: ev.Attempt,
-					Error:   ev.Error.Error(),
-				})
-			}
-		})
-	}
+	bridgeObserve(m, runner, extensions.Retry, func(ev RetryEvent) extensions.Event {
+		return extensions.RetryEvent{
+			Attempt: ev.Attempt,
+			Error:   ev.Error.Error(),
+		}
+	})
 
 	// --- PrepareStep hook ---
 	// Extension PrepareStep → SDK PrepareStep hook.
 	// Same pattern as ContextPrepare: convert LLMMessage ↔ ContextMessage.
 	if runner.HasHandlers(extensions.PrepareStep) {
 		m.OnPrepareStep(HookPriorityNormal, func(h PrepareStepHook) *PrepareStepResult {
-			// Convert LLM message slice to extension ContextMessage slice.
-			extMsgs := make([]extensions.ContextMessage, len(h.Messages))
-			for i, msg := range h.Messages {
-				var sb strings.Builder
-				for _, part := range msg.Content {
-					if tp, ok := part.(LLMTextPart); ok {
-						sb.WriteString(tp.Text)
-					}
-				}
-				extMsgs[i] = extensions.ContextMessage{
-					Index:   i,
-					Role:    string(msg.Role),
-					Content: sb.String(),
-				}
-			}
-
+			extMsgs := llmToContextMessages(h.Messages)
 			result, _ := runner.Emit(extensions.PrepareStepEvent{
 				StepNumber: h.StepNumber,
 				Messages:   extMsgs,
@@ -460,30 +344,71 @@ func (m *Kit) bridgeExtensions(runner *extensions.Runner) {
 			if !ok || r.Messages == nil {
 				return nil
 			}
-
-			// Rebuild LLM message slice from extension result.
-			rebuilt := make([]LLMMessage, 0, len(r.Messages))
-			for _, cm := range r.Messages {
-				if cm.Index >= 0 && cm.Index < len(h.Messages) {
-					rebuilt = append(rebuilt, h.Messages[cm.Index])
-				} else {
-					role := LLMRoleUser
-					switch cm.Role {
-					case "assistant":
-						role = LLMRoleAssistant
-					case "system":
-						role = LLMRoleSystem
-					case "tool":
-						role = LLMRoleTool
-					}
-					rebuilt = append(rebuilt, LLMMessage{
-						Role:    role,
-						Content: []LLMMessagePart{LLMTextPart{Text: cm.Content}},
-					})
-				}
-			}
-
-			return &PrepareStepResult{Messages: rebuilt}
+			return &PrepareStepResult{Messages: contextMessagesToLLM(r.Messages, h.Messages)}
 		})
 	}
 }
+
+// bridgeObserve subscribes to SDK events of type In and forwards them to the
+// extension runner as the event returned by conv. The subscription is only
+// registered when the runner has handlers for the given event kind.
+func bridgeObserve[In Event](m *Kit, runner *extensions.Runner, kind extensions.EventType, conv func(In) extensions.Event) {
+	if !runner.HasHandlers(kind) {
+		return
+	}
+	m.Subscribe(func(e Event) {
+		if ev, ok := e.(In); ok {
+			_, _ = runner.Emit(conv(ev))
+		}
+	})
+}
+
+// llmToContextMessages converts a slice of LLM messages to extension
+// ContextMessage values, extracting plain text from each message.
+func llmToContextMessages(msgs []LLMMessage) []extensions.ContextMessage {
+	extMsgs := make([]extensions.ContextMessage, len(msgs))
+	for i, msg := range msgs {
+		var sb strings.Builder
+		for _, part := range msg.Content {
+			if tp, ok := part.(LLMTextPart); ok {
+				sb.WriteString(tp.Text)
+			}
+		}
+		extMsgs[i] = extensions.ContextMessage{
+			Index:   i,
+			Role:    string(msg.Role),
+			Content: sb.String(),
+		}
+	}
+	return extMsgs
+}
+
+// contextMessagesToLLM rebuilds an LLM message slice from extension
+// ContextMessages. Messages with a valid index reuse the original from
+// originals; new messages injected by extensions are constructed from
+// role + text.
+func contextMessagesToLLM(cms []extensions.ContextMessage, originals []LLMMessage) []LLMMessage {
+	rebuilt := make([]LLMMessage, 0, len(cms))
+	for _, cm := range cms {
+		if cm.Index >= 0 && cm.Index < len(originals) {
+			// Reuse original message (preserves original role and content).
+			rebuilt = append(rebuilt, originals[cm.Index])
+		} else {
+			// New message injected by extension — construct from role + text.
+			role := LLMRoleUser
+			switch cm.Role {
+			case "assistant":
+				role = LLMRoleAssistant
+			case "system":
+				role = LLMRoleSystem
+			case "tool":
+				role = LLMRoleTool
+			}
+			rebuilt = append(rebuilt, LLMMessage{
+				Role:    role,
+				Content: []LLMMessagePart{LLMTextPart{Text: cm.Content}},
+			})
+		}
+	}
+	return rebuilt
+}

pkg/kit/kit.go

@@ -53,6 +53,14 @@ type Kit struct {
 	opts           *Options       // stored for reload operations (skills, etc.)
 	mcpConfig      *config.Config // loaded MCP/server config, shared with subagents
 
+	// v is this Kit instance's isolated configuration store. Each Kit owns its
+	// own *viper.Viper (constructed via viper.New) so that runtime config
+	// mutators (SetModel, SetThinkingLevel) and config reads do not clobber or
+	// observe state from other Kit instances in the same process. When the CLI
+	// constructs a Kit (Options.CLI != nil) this points at the process-global
+	// store so cobra flag bindings remain in effect.
+	v *viper.Viper
+
 	// hasCustomSystemPrompt is true when the user explicitly configured a
 	// system prompt (via --system-prompt flag, config file, or SDK option).
 	// When false, per-model system prompts from modelSettings/customModels
@@ -61,6 +69,11 @@ type Kit struct {
 	// systemPromptSource holds the raw configured value (file path or text)
 	// when hasCustomSystemPrompt is true; empty when the built-in default is in use.
 	systemPromptSource string
+	// basePrompt holds the resolved base system prompt text (post file-load,
+	// pre runtime-context composition) captured during New. Used by
+	// RefreshSystemPrompt to recompose after skills/context-file mutations.
+	// Protected by runtimeMu.
+	basePrompt string
 
 	// Hook registries — interception layer (see hooks.go).
 	beforeToolCall  *hookRegistry[BeforeToolCallHook, BeforeToolCallResult]
@@ -90,6 +103,12 @@ type Kit struct {
 		mu     sync.RWMutex
 	}
 
+	// runtimeMu protects contextFiles and skills against concurrent runtime
+	// mutations via AddSkill / RemoveSkill / AddContextFile etc. The fields
+	// are read by composeSystemPrompt and several other accessors, so all
+	// reads and writes after Kit construction must take this lock.
+	runtimeMu sync.RWMutex
+
 	// steerCh is a buffered channel used to inject steering messages into
 	// the running agent turn via the LLM library's PrepareStep. Created fresh for
 	// each generate() call and set to nil when idle. Protected by steerMu.
@@ -544,8 +563,8 @@ func (m *Kit) SetModel(ctx context.Context, modelString string) error {
 
 	// Build a provider config from current settings, overriding the model.
 	// Load system prompt properly (handles both file paths and inline content).
-	systemPrompt, _ := config.LoadSystemPrompt(viper.GetString("system-prompt"))
-	thinkingLevel := models.ParseThinkingLevel(viper.GetString("thinking-level"))
+	systemPrompt, _ := config.LoadSystemPrompt(m.v.GetString("system-prompt"))
+	thinkingLevel := models.ParseThinkingLevel(m.v.GetString("thinking-level"))
 
 	// Validate and adjust thinking level for the target model.
 	// Some models (e.g., OpenAI gpt-5.4) don't support "minimal" and require "none".
@@ -556,8 +575,8 @@ func (m *Kit) SetModel(ctx context.Context, modelString string) error {
 			if !models.IsValidThinkingLevelForModel(thinkingLevel, modelName) {
 				fallback := models.SuggestThinkingLevelFallback(thinkingLevel, modelName)
 				if fallback != models.ThinkingOff {
-					// Adjust the thinking level in viper so the change persists.
-					viper.Set("thinking-level", string(fallback))
+					// Adjust the thinking level in the instance store so the change persists.
+					m.v.Set("thinking-level", string(fallback))
 					thinkingLevel = fallback
 				}
 			}
@@ -569,35 +588,36 @@ func (m *Kit) SetModel(ctx context.Context, modelString string) error {
 	cfg := &models.ProviderConfig{
 		ModelString:    modelString,
 		SystemPrompt:   systemPrompt,
-		ProviderAPIKey: viper.GetString("provider-api-key"),
-		ProviderURL:    viper.GetString("provider-url"),
-		MaxTokens:      viper.GetInt("max-tokens"),
-		TLSSkipVerify:  viper.GetBool("tls-skip-verify"),
+		ProviderAPIKey: m.v.GetString("provider-api-key"),
+		ProviderURL:    m.v.GetString("provider-url"),
+		MaxTokens:      m.v.GetInt("max-tokens"),
+		TLSSkipVerify:  m.v.GetBool("tls-skip-verify"),
 		ThinkingLevel:  thinkingLevel,
 		DisableCaching: false, // Caching enabled by default, works with thinking
+		ConfigStore:    m.v,
 	}
 
 	// Only set generation parameter pointers when the user has explicitly
 	// provided a value. This leaves nil pointers for unset params, allowing
 	// per-model defaults (modelSettings / customModels params) to apply.
-	if viper.IsSet("temperature") {
-		v := float32(viper.GetFloat64("temperature"))
+	if m.v.IsSet("temperature") {
+		v := float32(m.v.GetFloat64("temperature"))
 		cfg.Temperature = &v
 	}
-	if viper.IsSet("top-p") {
-		v := float32(viper.GetFloat64("top-p"))
+	if m.v.IsSet("top-p") {
+		v := float32(m.v.GetFloat64("top-p"))
 		cfg.TopP = &v
 	}
-	if viper.IsSet("top-k") {
-		v := int32(viper.GetInt("top-k"))
+	if m.v.IsSet("top-k") {
+		v := int32(m.v.GetInt("top-k"))
 		cfg.TopK = &v
 	}
-	if viper.IsSet("frequency-penalty") {
-		v := float32(viper.GetFloat64("frequency-penalty"))
+	if m.v.IsSet("frequency-penalty") {
+		v := float32(m.v.GetFloat64("frequency-penalty"))
 		cfg.FrequencyPenalty = &v
 	}
-	if viper.IsSet("presence-penalty") {
-		v := float32(viper.GetFloat64("presence-penalty"))
+	if m.v.IsSet("presence-penalty") {
+		v := float32(m.v.GetFloat64("presence-penalty"))
 		cfg.PresencePenalty = &v
 	}
 
@@ -653,18 +673,25 @@ func (m *Kit) GetSystemPromptSource() string {
 // composeSystemPrompt takes a base system prompt and composes it with the
 // current runtime context: AGENTS.md content, skills metadata, and date/cwd.
 // This mirrors the composition done during Kit.New() initialization.
+// It acquires a read lock on runtimeMu while snapshotting contextFiles and
+// skills, so callers must not hold the write lock.
 func (m *Kit) composeSystemPrompt(basePrompt string) string {
 	cwd, _ := os.Getwd()
 	pb := skills.NewPromptBuilder(basePrompt)
 
+	m.runtimeMu.RLock()
+	contextFiles := append([]*ContextFile(nil), m.contextFiles...)
+	loadedSkills := append([]*skills.Skill(nil), m.skills...)
+	m.runtimeMu.RUnlock()
+
 	// Inject AGENTS.md content as project context.
-	for _, cf := range m.contextFiles {
+	for _, cf := range contextFiles {
 		pb.WithSection("", fmt.Sprintf("Instructions from: %s\n\n%s", cf.Path, cf.Content))
 	}
 
 	// Inject skills metadata.
-	if len(m.skills) > 0 {
-		pb.WithSkills(m.skills)
+	if len(loadedSkills) > 0 {
+		pb.WithSkills(loadedSkills)
 	}
 
 	// Append current date/time and working directory.
@@ -716,7 +743,7 @@ func (m *Kit) ReloadExtensions() error {
 	}
 
 	// Re-load from disk.
-	extraPaths := viper.GetStringSlice("extension")
+	extraPaths := m.v.GetStringSlice("extension")
 	loaded, err := extensions.LoadExtensions(extraPaths)
 	if err != nil {
 		return fmt.Errorf("reloading extensions: %w", err)
@@ -724,6 +751,7 @@ func (m *Kit) ReloadExtensions() error {
 
 	// Swap extensions on the runner (clears dynamic state).
 	m.extRunner.Reload(loaded)
+	m.extRunner.SetConfigStore(m.v)
 
 	// Update extension tools on the agent so the LLM sees changes.
 	if m.agent != nil {
@@ -762,7 +790,8 @@ func (m *Kit) ExecuteCompletion(ctx context.Context, req extensions.CompleteRequ
 		// Create a temporary provider for the requested model.
 		config := &models.ProviderConfig{
 			ModelString:   req.Model,
-			TLSSkipVerify: viper.GetBool("tls-skip-verify"),
+			TLSSkipVerify: m.v.GetBool("tls-skip-verify"),
+			ConfigStore:   m.v,
 		}
 		if req.MaxTokens > 0 {
 			config.MaxTokens = req.MaxTokens
@@ -848,37 +877,30 @@ func (m *Kit) ExecuteCompletion(ctx context.Context, req extensions.CompleteRequ
 // prompts, configuration, and behavior settings. All fields are optional
 // and will use CLI defaults if not specified.
 //
-// Global viper state warning:
-// Options are applied by [New] via [viper.Set] calls against viper's
-// process-global store. This store is shared with every downstream reader
-// (e.g. [Kit.SetModel], [Kit.GetThinkingLevel], BuildProviderConfig, and
-// any other code path that calls viper.Get*). Two consequences:
-//
-//  1. Kit instances are NOT isolated from each other within a single
-//     process. Values set by the second New() call overwrite the first,
-//     and any code that later reads viper will see the most recent Set.
-//  2. Fields left at the zero value do NOT clear prior viper state; they
-//     simply skip the viper.Set. Callers that need a clean slate between
-//     constructions should invoke viper.Reset() (the test suite uses a
-//     private resetViper() helper that wraps it) before the next New().
-//
-// Recommended usage: create one Kit per process, or reset viper between
-// constructions. Concurrent calls to New are serialized internally by
-// [viperInitMu], but that mutex does not prevent later viper reads (from
-// a different Kit) from observing mutated keys.
-//
-// TODO: refactor New to use a per-instance *viper.Viper (constructed via
-// viper.New()) so each Kit owns its own isolated config store and Options
-// no longer leak through the global singleton.
+// Config isolation: each [New] / [NewAgent] call constructs its own isolated
+// configuration store (via viper.New internally). Options are applied to that
+// per-instance store, so two Kits constructed in the same process do NOT share
+// or clobber each other's configuration. Runtime mutators ([Kit.SetModel],
+// [Kit.SetThinkingLevel]) and config readers ([Kit.GetThinkingLevel]) operate
+// only on the owning instance. Fields left at their zero value are simply not
+// applied; they fall through to the precedence chain (env → .kit.yml →
+// per-model defaults) resolved within the instance's own store.
 type Options struct {
 	Model        string // Override model (e.g., "anthropic/claude-sonnet-4-5-20250929")
 	SystemPrompt string // Override system prompt
 	ConfigFile   string // Override config file path
 	MaxSteps     int    // Override max steps (0 = use default)
-	Streaming    bool   // Enable streaming (default from config)
-	Quiet        bool   // Suppress debug output
-	Tools        []Tool // Custom tool set. If empty, AllTools() is used.
-	ExtraTools   []Tool // Additional tools added alongside core/MCP/extension tools.
+
+	// Streaming enables or disables streaming output. It is a pointer so the
+	// SDK can distinguish "unset" (nil) from an explicit choice, mirroring the
+	// sampling-parameter fields below. nil leaves streaming to the precedence
+	// chain (env → .kit.yml → default true); a non-nil value forces it. Prefer
+	// [WithStreaming] for the functional-options API.
+	Streaming *bool
+
+	Quiet      bool   // Suppress debug output
+	Tools      []Tool // Custom tool set. If empty, AllTools() is used.
+	ExtraTools []Tool // Additional tools added alongside core/MCP/extension tools.
 
 	// Generation parameters. These override the corresponding values from
 	// .kit.yml / KIT_* environment variables. Leaving a field at its
@@ -1151,40 +1173,40 @@ func InitTreeSession(opts *Options) (*session.TreeManager, error) {
 	return session.CreateTreeSession(sessionDir)
 }
 
-// viperInitMu serializes viper writes during [New]. Viper's global state
-// is not thread-safe, so concurrent calls (e.g. parallel subagent spawns)
-// must not overlap the Set/Get window. Note that this mutex only protects
-// the construction window — it does not isolate long-lived Kit instances
-// from each other. See the "Global viper state warning" on [Options].
-var viperInitMu sync.Mutex
-
 // New creates a Kit instance using the same initialization as the CLI.
 // It loads configuration, initializes MCP servers, creates the LLM model, and
 // sets up the agent for interaction. Returns an error if initialization fails.
 //
-// Global viper state warning: fields on [Options] are applied by calling
-// [viper.Set] on viper's process-global store. As a result, two Kits
-// constructed in the same process are NOT isolated: the second New
-// overwrites viper keys set by the first, and any downstream reader
-// (e.g. [Kit.SetModel], [Kit.GetThinkingLevel]) will observe the most
-// recent value. Callers that need multiple independent Kits should call
-// viper.Reset() between constructions, or avoid constructing more than
-// one Kit per process. Writes during New are serialized by [viperInitMu].
+// Config isolation: New constructs a per-instance configuration store (via
+// viper.New internally) and applies [Options] to it. Two Kits constructed in
+// the same process are therefore fully isolated — neither overwrites the
+// other's model, thinking level, or generation parameters, and runtime
+// mutators ([Kit.SetModel], [Kit.SetThinkingLevel]) only affect the owning
+// instance. This makes subagent spawning and multi-Kit embedding safe without
+// any external synchronization.
 //
-// TODO: refactor to use a per-call viper.New() instance so each Kit owns
-// its own isolated config store and Options stop leaking through the
-// global singleton.
+// CLI integration: when Options.CLI is non-nil the Kit shares the
+// process-global viper store instead of allocating a fresh one, so cobra flag
+// bindings established by the CLI remain in effect. SDK callers leave
+// Options.CLI nil and always get an isolated store.
+//
+// For an ergonomic functional-options front door, see [NewAgent].
 func New(ctx context.Context, opts *Options) (*Kit, error) {
 	if opts == nil {
 		opts = &Options{}
 	}
 
-	// All viper writes (SetSDKDefaults, InitConfig, Set calls, system-prompt
-	// composition) happen under viperInitMu. We also call BuildProviderConfig
-	// here — it's fast (just reads) — so we can capture the full config
-	// snapshot before releasing the lock. The expensive work (MCP loading,
-	// provider creation, session init) then runs outside the lock, allowing
-	// parallel subagent spawns to proceed concurrently.
+	// Construct this Kit's configuration store. SDK callers get a fresh,
+	// isolated *viper.Viper so concurrent constructions never clobber each
+	// other. The CLI (Options.CLI != nil) shares the process-global store so
+	// its cobra flag bindings and pre-loaded config remain visible.
+	var v *viper.Viper
+	if opts.CLI != nil {
+		v = viper.GetViper()
+	} else {
+		v = viper.New()
+	}
+
 	var (
 		providerConfig        *models.ProviderConfig
 		modelString           string
@@ -1194,86 +1216,93 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 		mcpConfig             *config.Config
 		debug                 bool
 		noExtensions          bool
+		disableCoreTools      bool
 		maxSteps              int
 		streaming             bool
 		hasCustomSystemPrompt bool
 		systemPromptSource    string
+		capturedBasePrompt    string
 	)
 
 	if err := func() error {
-		viperInitMu.Lock()
-		defer viperInitMu.Unlock()
+		// Set CLI-equivalent defaults on the instance store. When used as an
+		// SDK (without cobra), these defaults are not registered via flag bindings.
+		setSDKDefaults(v)
 
-		// Set CLI-equivalent defaults for viper. When used as an SDK (without
-		// cobra), these defaults are not registered via flag bindings.
-		setSDKDefaults()
-
-		// Initialize config (loads config files and env vars).
-		// Only initialize if not already done (e.g., by CLI's cobra.OnInitialize).
-		// Check if model is already set, which indicates config was loaded.
+		// Initialize config (loads config files and env vars) into the instance
+		// store. The CLI shares the process-global store, which cobra.OnInitialize
+		// has already populated, so re-running initConfig there is unnecessary;
+		// SDK callers get a fresh isolated store that must be loaded here.
+		// We key off opts.CLI (not a config value) because setSDKDefaults always
+		// seeds "model", which would otherwise mask an empty store.
 		// SkipConfig bypasses .kit.yml file loading (viper defaults and env vars still apply).
-		if !opts.SkipConfig && viper.GetString("model") == "" {
-			if err := InitConfig(opts.ConfigFile, false); err != nil {
+		if !opts.SkipConfig && opts.CLI == nil {
+			if err := initConfig(v, opts.ConfigFile, false); err != nil {
 				return fmt.Errorf("failed to initialize config: %w", err)
 			}
 		}
 
 		// Handle CLI debug mode.
 		if opts.Debug {
-			viper.Set("debug", true)
+			v.Set("debug", true)
 		}
 
-		// Override viper settings with options.
+		// Override instance settings with options.
 		if opts.Model != "" {
-			viper.Set("model", opts.Model)
+			v.Set("model", opts.Model)
 		}
 		if opts.SystemPrompt != "" {
-			viper.Set("system-prompt", opts.SystemPrompt)
+			v.Set("system-prompt", opts.SystemPrompt)
 		}
 		if opts.MaxSteps > 0 {
-			viper.Set("max-steps", opts.MaxSteps)
+			v.Set("max-steps", opts.MaxSteps)
+		}
+		// Only override streaming when the caller explicitly set it. Otherwise
+		// leave the precedence chain (env → config → default true) untouched so a
+		// zero-valued Options does not silently force stream=false.
+		if opts.Streaming != nil {
+			v.Set("stream", *opts.Streaming)
 		}
-		viper.Set("stream", opts.Streaming)
 
 		// Generation parameter overrides. Each Options field, when set,
-		// is pushed into viper here so the existing downstream code
-		// (BuildProviderConfig, SetModel, modelSettings lookups) picks
-		// it up uniformly. Pointer-typed sampling params use viper.Set
-		// only when non-nil so that nil means "leave provider/per-model
-		// default in place" (BuildProviderConfig keys off viper.IsSet).
+		// is pushed into the instance store here so the existing downstream
+		// code (BuildProviderConfig, SetModel, modelSettings lookups) picks
+		// it up uniformly. Pointer-typed sampling params use Set only when
+		// non-nil so that nil means "leave provider/per-model default in
+		// place" (BuildProviderConfig keys off IsSet).
 		if opts.MaxTokens > 0 {
-			viper.Set("max-tokens", opts.MaxTokens)
+			v.Set("max-tokens", opts.MaxTokens)
 		}
 		if opts.ThinkingLevel != "" {
-			viper.Set("thinking-level", opts.ThinkingLevel)
+			v.Set("thinking-level", opts.ThinkingLevel)
 		}
 		if opts.Temperature != nil {
-			viper.Set("temperature", *opts.Temperature)
+			v.Set("temperature", *opts.Temperature)
 		}
 		if opts.TopP != nil {
-			viper.Set("top-p", *opts.TopP)
+			v.Set("top-p", *opts.TopP)
 		}
 		if opts.TopK != nil {
-			viper.Set("top-k", *opts.TopK)
+			v.Set("top-k", *opts.TopK)
 		}
 		if opts.FrequencyPenalty != nil {
-			viper.Set("frequency-penalty", *opts.FrequencyPenalty)
+			v.Set("frequency-penalty", *opts.FrequencyPenalty)
 		}
 		if opts.PresencePenalty != nil {
-			viper.Set("presence-penalty", *opts.PresencePenalty)
+			v.Set("presence-penalty", *opts.PresencePenalty)
 		}
 
 		// Provider overrides. TLSSkipVerify only takes effect when true —
 		// callers wanting to force-disable should use the config file or
 		// env var instead.
 		if opts.ProviderAPIKey != "" {
-			viper.Set("provider-api-key", opts.ProviderAPIKey)
+			v.Set("provider-api-key", opts.ProviderAPIKey)
 		}
 		if opts.ProviderURL != "" {
-			viper.Set("provider-url", opts.ProviderURL)
+			v.Set("provider-url", opts.ProviderURL)
 		}
 		if opts.TLSSkipVerify {
-			viper.Set("tls-skip-verify", true)
+			v.Set("tls-skip-verify", true)
 		}
 
 		// Resolve working directory for context/skill discovery.
@@ -1304,7 +1333,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 		// explicitly set system-prompt, use the per-model prompt as the
 		// base instead of the global default.
 		{
-			rawPromptInput := viper.GetString("system-prompt")
+			rawPromptInput := v.GetString("system-prompt")
 
 			// Resolve a file path to its content so PromptBuilder receives the
 			// actual prompt text rather than a literal path string. Without this,
@@ -1329,12 +1358,12 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 			// Check for per-model system prompt override when no explicit
 			// global system-prompt was configured by the user.
 			if !userSetSystemPrompt {
-				modelStr := viper.GetString("model")
+				modelStr := v.GetString("model")
 				if modelStr != "" {
 					if mi := models.LookupModelForSettings(modelStr); mi != nil {
 						var perModelParams *models.GenerationParams
 						// modelSettings takes priority over custom model params.
-						if ms := models.LoadModelSettingsFromConfig(); ms != nil {
+						if ms := models.LoadModelSettingsFrom(v); ms != nil {
 							perModelParams = ms[modelStr]
 						}
 						if perModelParams == nil && mi.Params != nil {
@@ -1349,6 +1378,10 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 
 			pb := skills.NewPromptBuilder(basePrompt)
 
+			// Capture the resolved base prompt so RefreshSystemPrompt can
+			// recompose later after runtime skill/context-file mutations.
+			capturedBasePrompt = basePrompt
+
 			// Inject AGENTS.md content as project context.
 			for _, cf := range contextFiles {
 				pb.WithSection("", fmt.Sprintf("Instructions from: %s\n\n%s", cf.Path, cf.Content))
@@ -1365,41 +1398,42 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 				time.Now().Format("Monday, January 2, 2006, 3:04:05 PM MST"), cwd,
 			))
 
-			viper.Set("system-prompt", pb.Build())
+			v.Set("system-prompt", pb.Build())
 		}
 
-		// Snapshot all viper-derived values now, while the lock is held.
-		// BuildProviderConfig is fast (pure reads), so we do it here.
+		// Snapshot all instance-derived values now.
+		// BuildProviderConfig is fast (pure reads).
 		var pcErr error
-		providerConfig, _, pcErr = kitsetup.BuildProviderConfig()
+		providerConfig, _, pcErr = kitsetup.BuildProviderConfig(v)
 		if pcErr != nil {
 			return fmt.Errorf("failed to build provider config: %w", pcErr)
 		}
 
 		// SDK last-resort max-tokens floor. When nothing — Options, env,
 		// config, nor a per-model default — supplied a value, we land on
-		// zero here (viper.GetInt returns 0 for unset keys). Apply the
-		// SDK default directly on the struct rather than via viper so
-		// viper.IsSet("max-tokens") stays false: downstream right-sizing
+		// zero here (GetInt returns 0 for unset keys). Apply the
+		// SDK default directly on the struct rather than via the store so
+		// IsSet("max-tokens") stays false: downstream right-sizing
 		// can still raise this toward the model's known output ceiling,
 		// and per-model modelSettings[...].maxTokens can still win.
 		if providerConfig.MaxTokens == 0 && opts.MaxTokens == 0 {
 			providerConfig.MaxTokens = sdkDefaultMaxTokens
 		}
-		modelString = viper.GetString("model")
-		debug = viper.GetBool("debug")
-		noExtensions = opts.NoExtensions || viper.GetBool("no-extensions")
-		maxSteps = viper.GetInt("max-steps")
-		streaming = viper.GetBool("stream")
+		modelString = v.GetString("model")
+		debug = v.GetBool("debug")
+		noExtensions = opts.NoExtensions || v.GetBool("no-extensions")
+		disableCoreTools = opts.DisableCoreTools || v.GetBool("no-core-tools")
+		maxSteps = v.GetInt("max-steps")
+		streaming = v.GetBool("stream")
 
 		return nil
 	}(); err != nil {
 		return nil, err
 	}
-	// ---- viperInitMu released — heavy I/O below runs concurrently ----
+	// ---- config snapshot complete — heavy I/O below ----
 
 	// Load MCP configuration. Use pre-loaded config if provided directly,
-	// via CLI options, or load from viper as a last resort.
+	// via CLI options, or load from the instance store as a last resort.
 	if opts.MCPConfig != nil {
 		mcpConfig = opts.MCPConfig
 	} else if opts.CLI != nil && opts.CLI.MCPConfig != nil {
@@ -1407,7 +1441,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 	}
 	if mcpConfig == nil {
 		var err error
-		mcpConfig, err = config.LoadAndValidateConfig()
+		mcpConfig, err = config.LoadAndValidateConfigFrom(v)
 		if err != nil {
 			return nil, fmt.Errorf("failed to load MCP config: %w", err)
 		}
@@ -1446,7 +1480,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 		MCPConfig:         mcpConfig,
 		Quiet:             opts.Quiet,
 		CoreTools:         opts.Tools,
-		DisableCoreTools:  opts.DisableCoreTools,
+		DisableCoreTools:  disableCoreTools,
 		ExtraTools:        opts.ExtraTools,
 		ToolWrapper:       hookToolWrapper(beforeToolCall, afterToolResult),
 		ProviderConfig:    providerConfig,
@@ -1463,6 +1497,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 			timeout:         opts.MCPTaskTimeout,
 			progress:        opts.MCPTaskProgress,
 		}.toToolsConfig(),
+		Viper: v,
 	}
 
 	// Set up OAuth handler for remote MCP servers. The SDK does not create
@@ -1489,7 +1524,7 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 
 	if opts.CLI != nil {
 		setupOpts.ShowSpinner = opts.CLI.ShowSpinner
-		setupOpts.SpinnerFunc = opts.CLI.SpinnerFunc
+		setupOpts.SpinnerFunc = agent.SpinnerFunc(opts.CLI.SpinnerFunc)
 		setupOpts.UseBufferedLogger = opts.CLI.UseBufferedLogger
 		if opts.CLI.ProgressReaderFunc != nil {
 			providerConfig.ProgressReaderFunc = opts.CLI.ProgressReaderFunc
@@ -1532,8 +1567,10 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 		authHandler:           setupOpts.AuthHandler,
 		opts:                  opts,
 		mcpConfig:             mcpConfig,
+		v:                     v,
 		hasCustomSystemPrompt: hasCustomSystemPrompt,
 		systemPromptSource:    systemPromptSource,
+		basePrompt:            capturedBasePrompt,
 		beforeToolCall:        beforeToolCall,
 		afterToolResult:       afterToolResult,
 		beforeTurn:            beforeTurn,
@@ -1560,15 +1597,32 @@ func New(ctx context.Context, opts *Options) (*Kit, error) {
 	return k, nil
 }
 
-// GetContextFiles returns the context files (e.g. AGENTS.md) loaded during
-// initialisation. Returns nil if no context files were found.
+// GetContextFiles returns the context files (e.g. AGENTS.md) currently active
+// on this Kit instance. The returned slice is a snapshot — mutating it does
+// not affect Kit state. Returns nil when no context files are loaded.
 func (m *Kit) GetContextFiles() []*ContextFile {
-	return m.contextFiles
+	m.runtimeMu.RLock()
+	defer m.runtimeMu.RUnlock()
+	if len(m.contextFiles) == 0 {
+		return nil
+	}
+	out := make([]*ContextFile, len(m.contextFiles))
+	copy(out, m.contextFiles)
+	return out
 }
 
-// GetSkills returns the skills loaded during initialisation.
+// GetSkills returns the skills currently active on this Kit instance. The
+// returned slice is a snapshot — mutating it does not affect Kit state.
+// Returns nil when no skills are loaded.
 func (m *Kit) GetSkills() []*Skill {
-	return m.skills
+	m.runtimeMu.RLock()
+	defer m.runtimeMu.RUnlock()
+	if len(m.skills) == 0 {
+		return nil
+	}
+	out := make([]*Skill, len(m.skills))
+	copy(out, m.skills)
+	return out
 }
 
 // ---------------------------------------------------------------------------
@@ -1613,12 +1667,14 @@ func (m *Kit) expandSkillCommand(prompt string) string {
 
 	// Find the skill by name.
 	var skillPath string
+	m.runtimeMu.RLock()
 	for _, s := range m.skills {
 		if s.Name == name {
 			skillPath = s.Path
 			break
 		}
 	}
+	m.runtimeMu.RUnlock()
 	if skillPath == "" {
 		return prompt
 	}
@@ -1791,6 +1847,50 @@ type SubagentResult struct {
 	Elapsed time.Duration
 }
 
+// inheritProviderConfig copies the parent's effective provider/runtime
+// configuration from its isolated config store onto child Options. Used by
+// Kit.Subagent so the child — which owns a separate store and re-loads only
+// .kit.yml / KIT_* on its own — still observes provider credentials, the
+// thinking level, and sampler/token overrides the parent acquired via
+// programmatic Options or runtime setters (e.g. SetThinkingLevel).
+//
+// max-tokens and the sampling parameters are only propagated when the parent
+// explicitly set them (IsSet), preserving the tri-state precedence so per-model
+// defaults still apply on the child when the parent left them unset. A nil
+// child or store is a no-op.
+func inheritProviderConfig(child *Options, v *viper.Viper) {
+	if child == nil || v == nil {
+		return
+	}
+	child.ProviderAPIKey = v.GetString("provider-api-key")
+	child.ProviderURL = v.GetString("provider-url")
+	child.TLSSkipVerify = v.GetBool("tls-skip-verify")
+	child.ThinkingLevel = v.GetString("thinking-level")
+	if v.IsSet("max-tokens") {
+		child.MaxTokens = v.GetInt("max-tokens")
+	}
+	if v.IsSet("temperature") {
+		t := float32(v.GetFloat64("temperature"))
+		child.Temperature = &t
+	}
+	if v.IsSet("top-p") {
+		p := float32(v.GetFloat64("top-p"))
+		child.TopP = &p
+	}
+	if v.IsSet("top-k") {
+		k := int32(v.GetInt("top-k"))
+		child.TopK = &k
+	}
+	if v.IsSet("frequency-penalty") {
+		fp := float32(v.GetFloat64("frequency-penalty"))
+		child.FrequencyPenalty = &fp
+	}
+	if v.IsSet("presence-penalty") {
+		pp := float32(v.GetFloat64("presence-penalty"))
+		child.PresencePenalty = &pp
+	}
+}
+
 // Subagent spawns an in-process child Kit instance to perform a task. The
 // child gets its own session, event bus, and agent loop but shares the
 // parent's config (API keys, provider settings) and defaults to the parent's
@@ -1860,22 +1960,28 @@ func (m *Kit) Subagent(ctx context.Context, cfg SubagentConfig) (*SubagentResult
 	}
 
 	// Create child Kit instance. Pass the parent's loaded MCP config to
-	// avoid re-reading viper (which races with concurrent subagent spawns).
-	// Streaming must be explicitly enabled — Options.Streaming defaults to
-	// false, and New() unconditionally writes viper.Set("stream", opts.Streaming).
-	// Without this, the subagent would (a) pollute viper global state for
-	// other concurrent callers and (b) potentially hit provider-level
-	// differences (e.g. Anthropic non-streaming timeouts with extended
-	// thinking).
+	// avoid re-loading and re-validating config for the child.
+	// Streaming is enabled explicitly — without it, non-streaming can hit
+	// provider-level differences (e.g. Anthropic non-streaming timeouts with
+	// extended thinking). The child gets its own config store, so this does not
+	// affect any other concurrent caller.
+	streamOn := true
 	childOpts := &Options{
 		Model:        model,
 		SystemPrompt: systemPrompt,
 		Tools:        tools,
 		NoSession:    cfg.NoSession,
 		Quiet:        true,
-		Streaming:    true,
+		Streaming:    &streamOn,
 		MCPConfig:    m.mcpConfig,
 	}
+
+	// Inherit the parent's effective provider/runtime configuration. Since #40
+	// each Kit owns an isolated config store, so the child's New() only re-loads
+	// .kit.yml / KIT_* on its own — values the parent picked up from
+	// programmatic Options or runtime setters (e.g. SetThinkingLevel) would
+	// otherwise be lost.
+	inheritProviderConfig(childOpts, m.v)
 	// Propagate the parent's MCP task configuration so a child subagent
 	// invoking long-running MCP tools observes the same per-server modes,
 	// timeouts, and progress callback as the parent. Without this, child
@@ -2084,7 +2190,7 @@ func (m *Kit) generate(ctx context.Context, messages []fantasy.Message) (*agent.
 			}
 		},
 		OnStepUsage: func(inputTokens, outputTokens, cacheReadTokens, cacheCreationTokens int64) {
-			if viper.GetBool("debug") {
+			if m.v.GetBool("debug") {
 				log.Printf("DEBUG Kit.generate emitting StepUsageEvent: input=%d output=%d cacheRead=%d cacheCreate=%d",
 					inputTokens, outputTokens, cacheReadTokens, cacheCreationTokens,
 				)
@@ -2126,7 +2232,7 @@ func (m *Kit) generate(ctx context.Context, messages []fantasy.Message) (*agent.
 			})
 		},
 
-		// New callbacks for previously unwired Fantasy lifecycle events.
+		// New callbacks for previously unwired agent lifecycle events.
 		OnStepStart: func(stepNumber int) {
 			m.events.emit(StepStartEvent{StepNumber: stepNumber})
 		},
@@ -2580,7 +2686,7 @@ func (m *Kit) IsReasoningModel() bool {
 
 // GetThinkingLevel returns the current thinking level.
 func (m *Kit) GetThinkingLevel() string {
-	return viper.GetString("thinking-level")
+	return m.v.GetString("thinking-level")
 }
 
 // SetThinkingLevel changes the thinking level and recreates the agent with
@@ -2589,7 +2695,7 @@ func (m *Kit) GetThinkingLevel() string {
 // With message-level caching, both thinking and caching work together.
 // Caching reduces costs by 60-90% for repeated context.
 func (m *Kit) SetThinkingLevel(ctx context.Context, level string) error {
-	viper.Set("thinking-level", level)
+	m.v.Set("thinking-level", level)
 	// Recreate agent with new thinking config by re-running SetModel
 	// with the same model string. SetModel rebuilds the provider and
 	// passes the updated viper config (including thinking-level).

pkg/kit/kit_test.go

@@ -86,8 +86,8 @@ func TestNewWithGenerationOptions(t *testing.T) {
 		if got := host.MaxTokens(); got != want {
 			t.Errorf("Options.MaxTokens=%d did not propagate; Kit.MaxTokens()=%d", want, got)
 		}
-		if !viper.IsSet("max-tokens") {
-			t.Error("viper.IsSet(\"max-tokens\") should be true after MaxTokens override")
+		if !host.ConfigValueIsSetForTest("max-tokens") {
+			t.Error("max-tokens should be marked explicitly set on the instance store after MaxTokens override")
 		}
 	})
 
@@ -129,11 +129,11 @@ func TestNewWithGenerationOptions(t *testing.T) {
 		}
 		defer func() { _ = host.Close() }()
 
-		if !viper.IsSet("temperature") {
-			t.Fatal("viper.IsSet(\"temperature\") should be true after Temperature override")
+		if !host.ConfigValueIsSetForTest("temperature") {
+			t.Fatal("temperature should be marked explicitly set on the instance store after Temperature override")
 		}
-		if got := float32(viper.GetFloat64("temperature")); got != want {
-			t.Errorf("Options.Temperature=%v did not propagate; viper=%v", want, got)
+		if got := float32(host.ConfigFloatForTest("temperature")); got != want {
+			t.Errorf("Options.Temperature=%v did not propagate; instance store=%v", want, got)
 		}
 	})
 }
@@ -185,8 +185,8 @@ func TestNewPreservesIsSetSemantics(t *testing.T) {
 	// from SDK-side SetDefault/Set calls — which is exactly what this
 	// test is guarding against.
 	for _, k := range checkKeys {
-		if viper.IsSet(k) {
-			t.Errorf("viper.IsSet(%q) == true when no Options field set it "+
+		if host.ConfigValueIsSetForTest(k) {
+			t.Errorf("instance store reports %q explicitly set when no Options field set it "+
 				"(SDK defaults must not corrupt IsSet semantics)", k)
 		}
 	}
@@ -217,14 +217,14 @@ func TestNewWithProviderOptions(t *testing.T) {
 		}
 		defer func() { _ = host.Close() }()
 
-		if got := viper.GetString("provider-api-key"); got != apiKey {
-			t.Errorf("Options.ProviderAPIKey did not propagate to viper; got %q (len=%d)", got, len(got))
+		if got := host.ConfigStringForTest("provider-api-key"); got != apiKey {
+			t.Errorf("Options.ProviderAPIKey did not propagate to the instance store; got %q (len=%d)", got, len(got))
 		}
 	})
 
-	// Override precedence: even when viper already holds a different
-	// provider-api-key value (as it would if a config file or earlier
-	// Set() call populated one), Options.ProviderAPIKey must win.
+	// Override precedence: even when the process-global store already holds a
+	// different provider-api-key value, Options.ProviderAPIKey must win on the
+	// Kit's isolated store.
 	t.Run("Options override beats pre-existing viper state", func(t *testing.T) {
 		defer resetViper()
 
@@ -242,15 +242,16 @@ func TestNewWithProviderOptions(t *testing.T) {
 			ProviderAPIKey:   want,
 		})
 		// Creation may still fail if the model registry is strict, but
-		// we only care that the override reached viper before any
-		// provider handshake happened.
-		if host != nil {
-			defer func() { _ = host.Close() }()
+		// we only care that the override reached the instance store before
+		// any provider handshake happened.
+		if host == nil {
+			t.Fatalf("expected a Kit instance to inspect; got nil (err=%v)", err)
 		}
+		defer func() { _ = host.Close() }()
 		_ = err
 
-		if got := viper.GetString("provider-api-key"); got != want {
-			t.Errorf("Options.ProviderAPIKey did not override pre-existing viper value; got %q, want %q", got, want)
+		if got := host.ConfigStringForTest("provider-api-key"); got != want {
+			t.Errorf("Options.ProviderAPIKey did not override pre-existing value on the instance store; got %q, want %q", got, want)
 		}
 	})
 
@@ -270,7 +271,7 @@ func TestNewWithProviderOptions(t *testing.T) {
 		}
 		defer func() { _ = host.Close() }()
 
-		if got := viper.GetString("provider-url"); got != want {
+		if got := host.ConfigStringForTest("provider-url"); got != want {
 			t.Errorf("Options.ProviderURL did not propagate; got %q, want %q", got, want)
 		}
 	})
@@ -353,9 +354,9 @@ func TestNewSystemPromptFilePath(t *testing.T) {
 		t.Errorf("GetSystemPromptSource() = %q; want %q", got, want)
 	}
 
-	// The composed system prompt is written back to viper after PromptBuilder
-	// runs. It must contain the file's contents, not the file path.
-	composed := viper.GetString("system-prompt")
+	// The composed system prompt is written back to the instance store after
+	// PromptBuilder runs. It must contain the file's contents, not the file path.
+	composed := host.ConfigStringForTest("system-prompt")
 	if !strings.Contains(composed, promptContent) {
 		t.Errorf("composed system-prompt does not contain file contents\n  composed = %q\n  want substring = %q", composed, promptContent)
 	}
@@ -392,7 +393,7 @@ func TestNewSystemPromptInline(t *testing.T) {
 	if got := host.GetSystemPromptSource(); got != inline {
 		t.Errorf("GetSystemPromptSource() = %q; want %q", got, inline)
 	}
-	if composed := viper.GetString("system-prompt"); !strings.Contains(composed, inline) {
+	if composed := host.ConfigStringForTest("system-prompt"); !strings.Contains(composed, inline) {
 		t.Errorf("composed system-prompt missing inline content; got %q", composed)
 	}
 }

pkg/kit/mcp_tasks.go

@@ -98,6 +98,70 @@ type MCPTaskProgress struct {
 // dispatched on a goroutine.
 type MCPTaskProgressHandler func(MCPTaskProgress)
 
+// MCPTaskConfig configures task-aware MCP tools/call execution. All fields
+// are optional; the zero value disables progress callbacks and applies
+// sensible polling defaults inside the engine.
+//
+// For most consumers, the flat [Options] fields (`MCPTaskMode`,
+// `MCPTaskTTL`, `MCPTaskPollInterval`, `MCPTaskMaxPollInterval`,
+// `MCPTaskTimeout`, `MCPTaskProgress`) are the preferred entry point.
+// MCPTaskConfig is exposed for the low-level [AgentConfig] path.
+type MCPTaskConfig struct {
+	// PerServerMode overrides the per-server task mode resolved from
+	// [MCPServerConfig]. Keys are server names. Missing entries fall back
+	// to the configured value.
+	PerServerMode map[string]MCPTaskMode
+
+	// DefaultTTL is the TTL hint sent in TaskParams when augmenting a
+	// tools/call. Zero means omit the TTL — let the server pick its own.
+	DefaultTTL time.Duration
+
+	// PollInterval is the fallback interval between tasks/get requests
+	// when the server does not suggest one. Zero defaults to 1 second.
+	PollInterval time.Duration
+
+	// MaxPollInterval caps the polling interval. Zero defaults to 5 seconds.
+	MaxPollInterval time.Duration
+
+	// Timeout is the maximum wall-clock duration to wait for a task to
+	// reach a terminal state. Zero defaults to 15 minutes. Independent
+	// of the per-call context deadline; whichever fires first wins.
+	Timeout time.Duration
+
+	// Progress, if non-nil, receives every status transition observed by
+	// the polling loop.
+	Progress MCPTaskProgressHandler
+}
+
+// toToolsConfig converts the SDK-level [MCPTaskConfig] to the internal
+// tools-package representation. Keeps the dependency arrow internal-only.
+func (c MCPTaskConfig) toToolsConfig() tools.MCPTaskConfig {
+	cfg := tools.MCPTaskConfig{
+		DefaultTTL:      c.DefaultTTL,
+		PollInterval:    c.PollInterval,
+		MaxPollInterval: c.MaxPollInterval,
+		Timeout:         c.Timeout,
+	}
+	if len(c.PerServerMode) > 0 {
+		cfg.PerServerMode = make(map[string]tools.MCPTaskMode, len(c.PerServerMode))
+		for k, v := range c.PerServerMode {
+			cfg.PerServerMode[k] = tools.MCPTaskMode(v)
+		}
+	}
+	if c.Progress != nil {
+		h := c.Progress
+		cfg.Progress = func(p tools.MCPTaskProgress) {
+			h(MCPTaskProgress{
+				Server:  p.Server,
+				TaskID:  p.TaskID,
+				Status:  MCPTaskStatus(p.Status),
+				Message: p.Message,
+			})
+		}
+	}
+	return cfg
+}
+
 // mcpTaskOptions carries SDK consumer configuration into the agent setup.
 // Stored on Options as a single value so the public surface stays compact;
 // individual fields are exposed via WithMCP* builder functions.

pkg/kit/mcp_tasks_test.go

@@ -4,6 +4,8 @@ import (
 	"testing"
 	"time"
 
+	"github.com/spf13/viper"
+
 	"github.com/mark3labs/kit/internal/tools"
 )
 
@@ -163,3 +165,82 @@ func TestSubagentPropagatesMCPTaskOptions(t *testing.T) {
 	inheritMCPTaskOptions(&Options{}, nil)
 	inheritMCPTaskOptions(nil, parent)
 }
+
+// TestInheritProviderConfig verifies that Kit.Subagent's provider/runtime
+// config inheritance copies the parent's effective settings onto child
+// Options, and that the tri-state (IsSet) keys are only propagated when the
+// parent explicitly set them. Regression test for config loss after the
+// per-instance viper store isolation (#40).
+func TestInheritProviderConfig(t *testing.T) {
+	t.Run("explicit values propagate", func(t *testing.T) {
+		v := viper.New()
+		v.Set("provider-api-key", "sk-parent")
+		v.Set("provider-url", "https://proxy.internal/v1")
+		v.Set("tls-skip-verify", true)
+		v.Set("thinking-level", "high")
+		v.Set("max-tokens", 4321)
+		v.Set("temperature", 0.25)
+		v.Set("top-p", 0.9)
+		v.Set("top-k", 40)
+		v.Set("frequency-penalty", 0.1)
+		v.Set("presence-penalty", 0.2)
+
+		child := &Options{}
+		inheritProviderConfig(child, v)
+
+		if child.ProviderAPIKey != "sk-parent" {
+			t.Errorf("ProviderAPIKey = %q, want sk-parent", child.ProviderAPIKey)
+		}
+		if child.ProviderURL != "https://proxy.internal/v1" {
+			t.Errorf("ProviderURL = %q", child.ProviderURL)
+		}
+		if !child.TLSSkipVerify {
+			t.Error("TLSSkipVerify not propagated")
+		}
+		if child.ThinkingLevel != "high" {
+			t.Errorf("ThinkingLevel = %q, want high", child.ThinkingLevel)
+		}
+		if child.MaxTokens != 4321 {
+			t.Errorf("MaxTokens = %d, want 4321", child.MaxTokens)
+		}
+		if child.Temperature == nil || *child.Temperature != 0.25 {
+			t.Errorf("Temperature = %v, want 0.25", child.Temperature)
+		}
+		if child.TopP == nil || *child.TopP != 0.9 {
+			t.Errorf("TopP = %v, want 0.9", child.TopP)
+		}
+		if child.TopK == nil || *child.TopK != 40 {
+			t.Errorf("TopK = %v, want 40", child.TopK)
+		}
+		if child.FrequencyPenalty == nil || *child.FrequencyPenalty != 0.1 {
+			t.Errorf("FrequencyPenalty = %v, want 0.1", child.FrequencyPenalty)
+		}
+		if child.PresencePenalty == nil || *child.PresencePenalty != 0.2 {
+			t.Errorf("PresencePenalty = %v, want 0.2", child.PresencePenalty)
+		}
+	})
+
+	t.Run("unset tri-state keys stay unset", func(t *testing.T) {
+		// A store with no sampler / max-tokens keys must leave the child's
+		// pointers nil and MaxTokens zero so per-model defaults still apply.
+		v := viper.New()
+		child := &Options{}
+		inheritProviderConfig(child, v)
+
+		if child.MaxTokens != 0 {
+			t.Errorf("MaxTokens = %d, want 0 (unset)", child.MaxTokens)
+		}
+		if child.Temperature != nil || child.TopP != nil || child.TopK != nil ||
+			child.FrequencyPenalty != nil || child.PresencePenalty != nil {
+			t.Error("sampler pointers must stay nil when the parent did not set them")
+		}
+		if child.ThinkingLevel != "" {
+			t.Errorf("ThinkingLevel = %q, want empty", child.ThinkingLevel)
+		}
+	})
+
+	t.Run("nil child or store is a no-op", func(t *testing.T) {
+		inheritProviderConfig(nil, viper.New())
+		inheritProviderConfig(&Options{}, nil)
+	})
+}

pkg/kit/options.go

@@ -0,0 +1,88 @@
+package kit
+
+import "context"
+
+// Option configures a [Kit] created via [NewAgent]. Options are applied in
+// order to an [Options] value, so later options override earlier ones. The
+// type is a plain func(*Options), so callers can define their own options
+// without depending on any internal type.
+type Option func(*Options)
+
+// NewAgent creates a Kit using an ergonomic functional-options API. It is a
+// thin, additive front door over [New]: the supplied options are applied to a
+// fresh [Options] value which is then passed to [New]. For advanced
+// configuration not covered by the With* helpers (MCPConfig,
+// InProcessMCPServers, session backends, MCP task tuning, etc.) construct an
+// [Options] explicitly and call [New].
+//
+// Streaming defaults to enabled. Pass WithStreaming(false) to disable it.
+//
+// Example:
+//
+//	k, err := kit.NewAgent(ctx,
+//	    kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+//	    kit.WithSystemPrompt("You are a helpful assistant."),
+//	    kit.WithMaxTokens(8192),
+//	    kit.Ephemeral(),
+//	)
+func NewAgent(ctx context.Context, opts ...Option) (*Kit, error) {
+	// Streaming defaults to true for the ergonomic constructor — this is the
+	// natural expectation for interactive agents. WithStreaming(false) overrides it.
+	streamOn := true
+	o := &Options{Streaming: &streamOn}
+	for _, fn := range opts {
+		fn(o)
+	}
+	return New(ctx, o)
+}
+
+// WithModel sets the model in "provider/model" format
+// (e.g. "anthropic/claude-sonnet-4-5-20250929").
+func WithModel(m string) Option { return func(o *Options) { o.Model = m } }
+
+// WithSystemPrompt sets the system prompt. The value may be inline text or a
+// path to a file whose contents are loaded as the prompt.
+func WithSystemPrompt(p string) Option { return func(o *Options) { o.SystemPrompt = p } }
+
+// WithStreaming enables or disables streaming responses. [NewAgent] enables
+// streaming by default, so pass WithStreaming(false) to opt out.
+func WithStreaming(b bool) Option {
+	return func(o *Options) { o.Streaming = &b }
+}
+
+// WithMaxTokens sets the maximum output tokens per LLM response. A value of 0
+// lets the precedence chain (env → config → per-model → SDK floor) resolve a
+// value; a non-zero value pins it and suppresses automatic right-sizing.
+func WithMaxTokens(n int) Option { return func(o *Options) { o.MaxTokens = n } }
+
+// WithThinkingLevel sets the reasoning effort for models that support extended
+// thinking. Valid values: "off", "none", "minimal", "low", "medium", "high".
+// An empty string lets the precedence chain resolve a level.
+func WithThinkingLevel(level string) Option { return func(o *Options) { o.ThinkingLevel = level } }
+
+// WithTools sets the agent's tool set, replacing the default core tools. When
+// no tools are provided the default set is used.
+func WithTools(t ...Tool) Option { return func(o *Options) { o.Tools = t } }
+
+// WithExtraTools adds tools alongside the core/MCP/extension tools rather than
+// replacing them.
+func WithExtraTools(t ...Tool) Option { return func(o *Options) { o.ExtraTools = t } }
+
+// WithProviderAPIKey overrides the API key used to authenticate with the model
+// provider.
+func WithProviderAPIKey(key string) Option { return func(o *Options) { o.ProviderAPIKey = key } }
+
+// WithProviderURL overrides the provider endpoint URL. Useful for
+// OpenAI-compatible proxies (LiteLLM, vLLM, Azure OpenAI, etc.).
+func WithProviderURL(url string) Option { return func(o *Options) { o.ProviderURL = url } }
+
+// WithConfigFile sets an explicit config file path, overriding the default
+// .kit.yml search.
+func WithConfigFile(path string) Option { return func(o *Options) { o.ConfigFile = path } }
+
+// WithDebug enables SDK debug logging.
+func WithDebug() Option { return func(o *Options) { o.Debug = true } }
+
+// Ephemeral configures an in-memory session with no persistence (equivalent to
+// Options.NoSession = true).
+func Ephemeral() Option { return func(o *Options) { o.NoSession = true } }

pkg/kit/runtime_skills_context_test.go

@@ -0,0 +1,342 @@
+package kit
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"testing"
+
+	"github.com/mark3labs/kit/internal/agent"
+	"github.com/mark3labs/kit/internal/skills"
+)
+
+// TestAddSkill_AddsAndDeduplicates verifies that AddSkill registers new skills
+// and that re-adding a skill with the same Name replaces the existing entry
+// rather than appending a duplicate. agent is nil in these tests; the method
+// must still mutate the in-memory state and tolerate the absent agent.
+func TestAddSkill_AddsAndDeduplicates(t *testing.T) {
+	k := &Kit{basePrompt: "base"}
+
+	if err := k.AddSkill(&skills.Skill{Name: "alpha", Content: "first"}); err != nil {
+		t.Fatalf("AddSkill alpha: %v", err)
+	}
+	if err := k.AddSkill(&skills.Skill{Name: "beta", Content: "second"}); err != nil {
+		t.Fatalf("AddSkill beta: %v", err)
+	}
+	got := k.GetSkills()
+	if len(got) != 2 {
+		t.Fatalf("expected 2 skills, got %d", len(got))
+	}
+
+	// Re-adding alpha with new content must replace, not duplicate.
+	if err := k.AddSkill(&skills.Skill{Name: "alpha", Content: "replaced"}); err != nil {
+		t.Fatalf("AddSkill alpha replace: %v", err)
+	}
+	got = k.GetSkills()
+	if len(got) != 2 {
+		t.Fatalf("expected 2 skills after replace, got %d", len(got))
+	}
+	for _, s := range got {
+		if s.Name == "alpha" && s.Content != "replaced" {
+			t.Errorf("alpha content = %q; want %q", s.Content, "replaced")
+		}
+	}
+}
+
+// TestAddSkill_Validation rejects nil skills and unnamed skills with errors
+// instead of corrupting state.
+func TestAddSkill_Validation(t *testing.T) {
+	k := &Kit{}
+	if err := k.AddSkill(nil); err == nil {
+		t.Error("expected error for nil skill")
+	}
+	if err := k.AddSkill(&skills.Skill{Content: "x"}); err == nil {
+		t.Error("expected error for unnamed skill")
+	}
+	if got := k.GetSkills(); got != nil {
+		t.Errorf("skills list mutated after invalid AddSkill calls: %#v", got)
+	}
+}
+
+// TestRemoveSkill verifies removal and the false return for misses.
+func TestRemoveSkill(t *testing.T) {
+	k := &Kit{}
+	_ = k.AddSkill(&skills.Skill{Name: "alpha"})
+	_ = k.AddSkill(&skills.Skill{Name: "beta"})
+
+	if removed := k.RemoveSkill("missing"); removed {
+		t.Error("RemoveSkill(missing) = true; want false")
+	}
+	if removed := k.RemoveSkill("alpha"); !removed {
+		t.Error("RemoveSkill(alpha) = false; want true")
+	}
+	got := k.GetSkills()
+	if len(got) != 1 || got[0].Name != "beta" {
+		t.Errorf("remaining skills = %#v; want [beta]", got)
+	}
+}
+
+// TestSetSkills replaces the entire set and validates input.
+func TestSetSkills(t *testing.T) {
+	k := &Kit{}
+	_ = k.AddSkill(&skills.Skill{Name: "alpha"})
+
+	err := k.SetSkills([]*skills.Skill{
+		{Name: "one"},
+		{Name: "two"},
+		{Name: "three"},
+	})
+	if err != nil {
+		t.Fatalf("SetSkills: %v", err)
+	}
+	if got := k.GetSkills(); len(got) != 3 {
+		t.Errorf("expected 3 skills, got %d", len(got))
+	}
+
+	// Invalid entry rejects the whole batch.
+	bad := []*skills.Skill{{Name: "ok"}, nil}
+	if err := k.SetSkills(bad); err == nil {
+		t.Error("expected error when batch contains nil")
+	}
+	// State unchanged after rejected batch.
+	if got := k.GetSkills(); len(got) != 3 {
+		t.Errorf("skills mutated by rejected SetSkills batch: len=%d", len(got))
+	}
+
+	// Empty slice clears.
+	if err := k.SetSkills(nil); err != nil {
+		t.Fatalf("SetSkills(nil): %v", err)
+	}
+	if got := k.GetSkills(); got != nil {
+		t.Errorf("expected nil skills after clear; got %#v", got)
+	}
+}
+
+// TestLoadAndAddSkill round-trips a skill file from disk.
+func TestLoadAndAddSkill(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "demo.md")
+	body := "---\nname: demo\ndescription: demo skill\n---\nhello world"
+	if err := os.WriteFile(path, []byte(body), 0o644); err != nil {
+		t.Fatalf("write skill file: %v", err)
+	}
+
+	k := &Kit{}
+	s, err := k.LoadAndAddSkill(path)
+	if err != nil {
+		t.Fatalf("LoadAndAddSkill: %v", err)
+	}
+	if s.Name != "demo" {
+		t.Errorf("loaded skill Name = %q; want demo", s.Name)
+	}
+	if got := k.GetSkills(); len(got) != 1 {
+		t.Errorf("expected 1 skill registered, got %d", len(got))
+	}
+}
+
+// TestAddContextFile_DeduplicatesByPath confirms identical paths replace
+// rather than duplicate.
+func TestAddContextFile_DeduplicatesByPath(t *testing.T) {
+	k := &Kit{}
+	if err := k.AddContextFile(&ContextFile{Path: "/a/AGENTS.md", Content: "v1"}); err != nil {
+		t.Fatalf("AddContextFile: %v", err)
+	}
+	if err := k.AddContextFile(&ContextFile{Path: "/b/AGENTS.md", Content: "vB"}); err != nil {
+		t.Fatalf("AddContextFile: %v", err)
+	}
+	if err := k.AddContextFile(&ContextFile{Path: "/a/AGENTS.md", Content: "v2"}); err != nil {
+		t.Fatalf("AddContextFile replace: %v", err)
+	}
+
+	got := k.GetContextFiles()
+	if len(got) != 2 {
+		t.Fatalf("expected 2 context files, got %d", len(got))
+	}
+	for _, cf := range got {
+		if cf.Path == "/a/AGENTS.md" && cf.Content != "v2" {
+			t.Errorf("/a/AGENTS.md content = %q; want v2", cf.Content)
+		}
+	}
+}
+
+// TestAddContextFile_Validation rejects nil and unpathed entries.
+func TestAddContextFile_Validation(t *testing.T) {
+	k := &Kit{}
+	if err := k.AddContextFile(nil); err == nil {
+		t.Error("expected error for nil context file")
+	}
+	if err := k.AddContextFile(&ContextFile{Content: "x"}); err == nil {
+		t.Error("expected error for empty path")
+	}
+}
+
+// TestRemoveContextFile_Behavior verifies remove returns true on hit and
+// false on miss without mutating state on a miss.
+func TestRemoveContextFile_Behavior(t *testing.T) {
+	k := &Kit{}
+	_ = k.AddContextFile(&ContextFile{Path: "/a", Content: "x"})
+	_ = k.AddContextFile(&ContextFile{Path: "/b", Content: "y"})
+
+	if removed := k.RemoveContextFile("/missing"); removed {
+		t.Error("RemoveContextFile(missing) = true; want false")
+	}
+	if removed := k.RemoveContextFile("/a"); !removed {
+		t.Error("RemoveContextFile(/a) = false; want true")
+	}
+	got := k.GetContextFiles()
+	if len(got) != 1 || got[0].Path != "/b" {
+		t.Errorf("remaining = %#v; want [/b]", got)
+	}
+}
+
+// TestSetContextFiles replaces and validates batch input.
+func TestSetContextFiles(t *testing.T) {
+	k := &Kit{}
+	_ = k.AddContextFile(&ContextFile{Path: "/seed", Content: "old"})
+
+	err := k.SetContextFiles([]*ContextFile{
+		{Path: "/x", Content: "x"},
+		{Path: "/y", Content: "y"},
+	})
+	if err != nil {
+		t.Fatalf("SetContextFiles: %v", err)
+	}
+	if got := k.GetContextFiles(); len(got) != 2 {
+		t.Errorf("expected 2 context files, got %d", len(got))
+	}
+
+	bad := []*ContextFile{{Path: "/ok"}, {Path: ""}}
+	if err := k.SetContextFiles(bad); err == nil {
+		t.Error("expected error for empty path in batch")
+	}
+	if got := k.GetContextFiles(); len(got) != 2 {
+		t.Errorf("state mutated by rejected batch: len=%d", len(got))
+	}
+
+	if err := k.SetContextFiles(nil); err != nil {
+		t.Fatalf("SetContextFiles(nil): %v", err)
+	}
+	if got := k.GetContextFiles(); got != nil {
+		t.Errorf("expected nil after clear; got %#v", got)
+	}
+}
+
+// TestLoadAndAddContextFile reads from disk and registers the context file.
+func TestLoadAndAddContextFile(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "AGENTS.md")
+	const content = "# Agent rules\nuse the new lint config"
+	if err := os.WriteFile(path, []byte(content), 0o644); err != nil {
+		t.Fatalf("write file: %v", err)
+	}
+
+	k := &Kit{}
+	cf, err := k.LoadAndAddContextFile(path)
+	if err != nil {
+		t.Fatalf("LoadAndAddContextFile: %v", err)
+	}
+	if !strings.HasSuffix(cf.Path, "AGENTS.md") {
+		t.Errorf("Path = %q; want suffix AGENTS.md", cf.Path)
+	}
+	if !strings.Contains(cf.Content, "use the new lint config") {
+		t.Errorf("Content missing expected body: %q", cf.Content)
+	}
+	got := k.GetContextFiles()
+	if len(got) != 1 {
+		t.Fatalf("expected 1 context file, got %d", len(got))
+	}
+}
+
+// TestAddContextFileContent registers an in-memory context blob.
+func TestAddContextFileContent(t *testing.T) {
+	k := &Kit{}
+	cf, err := k.AddContextFileContent("session://user-123/AGENTS.md", "always greet in French")
+	if err != nil {
+		t.Fatalf("AddContextFileContent: %v", err)
+	}
+	if cf.Path != "session://user-123/AGENTS.md" {
+		t.Errorf("Path = %q", cf.Path)
+	}
+	if cf.Content != "always greet in French" {
+		t.Errorf("Content = %q", cf.Content)
+	}
+}
+
+// TestComposeSystemPrompt_IncludesSkillsAndContext verifies that runtime
+// mutations actually flow into the composed system prompt that the agent
+// would receive.
+func TestComposeSystemPrompt_IncludesSkillsAndContext(t *testing.T) {
+	k := &Kit{basePrompt: "BASE-PROMPT-MARKER"}
+
+	if err := k.AddContextFile(&ContextFile{
+		Path:    "/proj/AGENTS.md",
+		Content: "CTX-MARKER-OK",
+	}); err != nil {
+		t.Fatalf("AddContextFile: %v", err)
+	}
+	if err := k.AddSkill(&skills.Skill{
+		Name:        "greeter",
+		Description: "SKILL-DESC-MARKER",
+		Content:     "do greetings",
+		Path:        "/skills/greeter.md",
+	}); err != nil {
+		t.Fatalf("AddSkill: %v", err)
+	}
+
+	composed := k.composeSystemPrompt(k.basePrompt)
+	for _, want := range []string{
+		"BASE-PROMPT-MARKER",
+		"CTX-MARKER-OK",
+		"/proj/AGENTS.md",
+		"greeter",
+		"SKILL-DESC-MARKER",
+	} {
+		if !strings.Contains(composed, want) {
+			t.Errorf("composed prompt missing %q\n--- composed ---\n%s", want, composed)
+		}
+	}
+
+	// Removing the skill should remove its marker from the next composition.
+	k.RemoveSkill("greeter")
+	composed = k.composeSystemPrompt(k.basePrompt)
+	if strings.Contains(composed, "SKILL-DESC-MARKER") {
+		t.Errorf("composed prompt still contains removed skill description:\n%s", composed)
+	}
+}
+
+// TestRuntimeMutations_AreThreadSafe stresses the mutation API from multiple
+// goroutines to surface data races under `go test -race`.
+func TestRuntimeMutations_AreThreadSafe(t *testing.T) {
+	// Use a non-nil agent so applyComposedSystemPrompt actually invokes
+	// agent.SetSystemPrompt (a no-op agent is fine — we only need the
+	// systemPrompt mutation + fantasy rebuild path to run concurrently so
+	// -race can observe any unsynchronized writes).
+	k := &Kit{basePrompt: "base", agent: &agent.Agent{}}
+	var wg sync.WaitGroup
+	const goroutines = 8
+	const iterations = 50
+
+	for g := range goroutines {
+		wg.Add(1)
+		go func(id int) {
+			defer wg.Done()
+			for range iterations {
+				_ = k.AddSkill(&skills.Skill{
+					Name:    "skill",
+					Content: "content",
+				})
+				_ = k.AddContextFile(&ContextFile{
+					Path:    "/shared/AGENTS.md",
+					Content: "shared",
+				})
+				_ = k.GetSkills()
+				_ = k.GetContextFiles()
+				_ = k.composeSystemPrompt("base")
+				k.RemoveSkill("skill")
+				k.RemoveContextFile("/shared/AGENTS.md")
+			}
+		}(g)
+	}
+	wg.Wait()
+}

pkg/kit/skills.go

@@ -139,13 +139,150 @@ func (m *Kit) ClearSkillCache() {
 }
 
 // ReloadSkills re-discovers skills from disk, replacing the current set.
-// This is called by file watchers when skill files change.
+// This is called by file watchers when skill files change. The system prompt
+// is recomposed and applied to the running agent so subsequent turns see the
+// new skill set.
 func (m *Kit) ReloadSkills() error {
 	newSkills, err := loadSkills(m.opts)
 	if err != nil {
 		return fmt.Errorf("reloading skills: %w", err)
 	}
+	m.runtimeMu.Lock()
 	m.skills = newSkills
+	m.runtimeMu.Unlock()
 	m.ClearSkillCache()
+	m.applyComposedSystemPrompt()
 	return nil
 }
+
+// ---------------------------------------------------------------------------
+// Runtime skill management (Issue #36)
+// ---------------------------------------------------------------------------
+//
+// The methods below let SDK consumers (chatbot hosts, multi-tenant agents)
+// mutate the active skill set after Kit construction. Each mutation recomposes
+// the system prompt and applies it to the underlying agent so the LLM sees
+// the new skill metadata on its next turn.
+
+// AddSkill registers a single skill on this Kit instance. The skill object
+// can be built programmatically (no file on disk required) — only Name and
+// Content are mandatory. If a skill with the same Name is already loaded the
+// new skill replaces it. Returns an error when skill is nil or has an empty
+// name.
+//
+// After mutation the system prompt is recomposed and applied to the running
+// agent so the next turn sees the updated skill metadata. AddSkill is safe to
+// call from any goroutine.
+func (m *Kit) AddSkill(skill *Skill) error {
+	if skill == nil {
+		return fmt.Errorf("AddSkill: skill is nil")
+	}
+	if skill.Name == "" {
+		return fmt.Errorf("AddSkill: skill name is required")
+	}
+
+	m.runtimeMu.Lock()
+	replaced := false
+	for i, s := range m.skills {
+		if s.Name == skill.Name {
+			m.skills[i] = skill
+			replaced = true
+			break
+		}
+	}
+	if !replaced {
+		m.skills = append(m.skills, skill)
+	}
+	m.runtimeMu.Unlock()
+
+	m.ClearSkillCache()
+	m.applyComposedSystemPrompt()
+	return nil
+}
+
+// LoadAndAddSkill loads a skill from a filesystem path (single .md/.txt file)
+// and adds it via [Kit.AddSkill]. Returns the loaded skill on success.
+func (m *Kit) LoadAndAddSkill(path string) (*Skill, error) {
+	s, err := skills.LoadSkill(path)
+	if err != nil {
+		return nil, fmt.Errorf("LoadAndAddSkill: %w", err)
+	}
+	if err := m.AddSkill(s); err != nil {
+		return nil, err
+	}
+	return s, nil
+}
+
+// RemoveSkill removes the named skill from this Kit instance and recomposes
+// the system prompt. Returns true when a skill with that name was found and
+// removed, false otherwise.
+func (m *Kit) RemoveSkill(name string) bool {
+	m.runtimeMu.Lock()
+	found := false
+	for i, s := range m.skills {
+		if s.Name == name {
+			m.skills = append(m.skills[:i], m.skills[i+1:]...)
+			found = true
+			break
+		}
+	}
+	m.runtimeMu.Unlock()
+
+	if !found {
+		return false
+	}
+	m.ClearSkillCache()
+	m.applyComposedSystemPrompt()
+	return true
+}
+
+// SetSkills replaces the active skill set with the provided slice. Pass nil
+// or an empty slice to remove all skills. The system prompt is recomposed and
+// applied. Skills with empty names are rejected and no mutation is performed.
+func (m *Kit) SetSkills(skillList []*Skill) error {
+	// Validate first so a bad input doesn't partially mutate state.
+	for i, s := range skillList {
+		if s == nil {
+			return fmt.Errorf("SetSkills: skill at index %d is nil", i)
+		}
+		if s.Name == "" {
+			return fmt.Errorf("SetSkills: skill at index %d has empty name", i)
+		}
+	}
+
+	copied := make([]*Skill, len(skillList))
+	copy(copied, skillList)
+
+	m.runtimeMu.Lock()
+	m.skills = copied
+	m.runtimeMu.Unlock()
+
+	m.ClearSkillCache()
+	m.applyComposedSystemPrompt()
+	return nil
+}
+
+// applyComposedSystemPrompt recomposes the system prompt from the captured
+// base prompt + current contextFiles + current skills + date/cwd, and pushes
+// the result onto the underlying agent. No-op when the agent is unset (i.e.
+// during construction).
+func (m *Kit) applyComposedSystemPrompt() {
+	if m.agent == nil {
+		return
+	}
+	m.runtimeMu.RLock()
+	base := m.basePrompt
+	m.runtimeMu.RUnlock()
+	composed := m.composeSystemPrompt(base)
+	m.agent.SetSystemPrompt(composed)
+}
+
+// RefreshSystemPrompt manually recomposes the system prompt from the current
+// skills and context files and applies it to the agent. Call this after a
+// batch of low-level mutations or to force a re-render of the date/cwd
+// section. Most callers don't need to invoke this directly because
+// AddSkill, RemoveSkill, SetSkills, AddContextFile, RemoveContextFile, and
+// SetContextFiles all refresh automatically.
+func (m *Kit) RefreshSystemPrompt() {
+	m.applyComposedSystemPrompt()
+}

pkg/kit/types.go

@@ -11,6 +11,7 @@ import (
 	"github.com/mark3labs/kit/internal/message"
 	"github.com/mark3labs/kit/internal/models"
 	"github.com/mark3labs/kit/internal/session"
+	"github.com/mark3labs/kit/internal/tools"
 	"github.com/mark3labs/mcp-go/client/transport"
 	"github.com/mark3labs/mcp-go/server"
 )
@@ -75,25 +76,151 @@ type Config = config.Config
 // local (stdio) and remote (StreamableHTTP/SSE) server types.
 type MCPServerConfig = config.MCPServerConfig
 
-// ==== Agent Types (internal/agent/) ====
+// ==== Agent Types ====
 
-// AgentConfig holds configuration options for creating a new Agent.
-type AgentConfig = agent.AgentConfig
+// DebugLogger is an SDK-owned interface for low-level debug logging from
+// the engine and MCP tool plumbing. Implementations must be safe for
+// concurrent use.
+//
+// Most consumers do not need to provide one; pass [Options.Debug] = true
+// to use the default logger. DebugLogger is exposed for the low-level
+// [AgentConfig] path and for embedders that want to route debug output
+// into their own logging system.
+type DebugLogger interface {
+	// LogDebug records a single debug message. Implementations may drop,
+	// buffer, or render the message however they choose.
+	LogDebug(message string)
+	// IsDebugEnabled reports whether debug logging is active. Callers may
+	// check this before doing expensive formatting work.
+	IsDebugEnabled() bool
+}
 
-type (
-	// ToolCallHandler is a function type for handling tool calls as they happen.
-	ToolCallHandler = agent.ToolCallHandler
-	// ToolExecutionHandler is a function type for handling tool execution start/end events.
-	ToolExecutionHandler = agent.ToolExecutionHandler
-	// ToolResultHandler is a function type for handling tool results.
-	ToolResultHandler = agent.ToolResultHandler
-	// ResponseHandler is a function type for handling LLM responses.
-	ResponseHandler = agent.ResponseHandler
-	// StreamingResponseHandler is a function type for handling streaming LLM responses.
-	StreamingResponseHandler = agent.StreamingResponseHandler
-	// ToolCallContentHandler is a function type for handling content that accompanies tool calls.
-	ToolCallContentHandler = agent.ToolCallContentHandler
-)
+// AgentConfig holds configuration options for constructing an agent at the
+// SDK boundary. All fields use SDK-owned types, so consumers can populate
+// this struct without importing any underlying LLM-provider package.
+//
+// For most use cases, prefer the high-level [New] entry point with
+// [Options]. AgentConfig is exposed for advanced consumers that need
+// direct access to the lower-level agent configuration shape.
+type AgentConfig struct {
+	// ModelConfig holds the LLM provider configuration. A nil value means
+	// that the default provider/model resolution will be used.
+	ModelConfig *ProviderConfig
+
+	// MCPConfig describes any MCP servers whose tools should be loaded
+	// alongside core tools.
+	MCPConfig *Config
+
+	// SystemPrompt is the system prompt sent to the LLM.
+	SystemPrompt string
+
+	// MaxSteps caps the number of LLM iterations per turn. A value of
+	// zero means no cap is applied at this layer.
+	MaxSteps int
+
+	// StreamingEnabled controls whether the agent streams responses.
+	StreamingEnabled bool
+
+	// AuthHandler handles OAuth authorization for remote MCP servers.
+	// When nil, remote MCP servers requiring OAuth will fail to connect.
+	AuthHandler MCPAuthHandler
+
+	// TokenStoreFactory, if non-nil, creates a custom token store for each
+	// remote MCP server's OAuth tokens. When nil, the default file-based
+	// token store is used.
+	TokenStoreFactory MCPTokenStoreFactory
+
+	// CoreTools overrides the default core tool set. If empty, [AllTools]
+	// is used. Provide a custom tool set (e.g. [CodingTools] or tools
+	// built with a custom WorkDir) to scope agent capabilities.
+	CoreTools []Tool
+
+	// DisableCoreTools, when true, prevents loading any core tools.
+	// Combined with empty CoreTools this yields a chat-only agent with
+	// no built-in tools.
+	DisableCoreTools bool
+
+	// ExtraTools are additional tools loaded alongside core and MCP tools.
+	ExtraTools []Tool
+
+	// ToolWrapper, if non-nil, wraps the combined tool list before it is
+	// handed to the LLM. Used to intercept tool calls or results.
+	ToolWrapper func([]Tool) []Tool
+
+	// OnMCPServerLoaded, if non-nil, is invoked once for each MCP server
+	// when its tools have finished loading (or failed). Called from a
+	// background goroutine.
+	OnMCPServerLoaded func(serverName string, toolCount int, err error)
+
+	// DebugLogger receives low-level debug output from the engine and the
+	// MCP tool plumbing. Nil means no debug output is emitted at this
+	// layer (regardless of [Options.Debug], which feeds the higher-level
+	// [New] entry point). Pass an implementation here when wiring a custom
+	// logger through the lower-level AgentConfig path.
+	DebugLogger DebugLogger
+
+	// MCPTaskConfig configures task-aware MCP tools/call execution — mode
+	// overrides, polling intervals, timeouts, and the progress handler.
+	// The zero value preserves historical synchronous-only behaviour for
+	// any server that didn't advertise task support during initialize.
+	MCPTaskConfig MCPTaskConfig
+}
+
+// toInternal converts an AgentConfig to its internal representation.
+// Slice and function fields convert without allocation because [Tool]
+// is a type alias for the underlying LLM-tool type.
+func (c *AgentConfig) toInternal() *agent.AgentConfig {
+	if c == nil {
+		return nil
+	}
+	out := &agent.AgentConfig{
+		ModelConfig:       c.ModelConfig,
+		MCPConfig:         c.MCPConfig,
+		SystemPrompt:      c.SystemPrompt,
+		MaxSteps:          c.MaxSteps,
+		StreamingEnabled:  c.StreamingEnabled,
+		CoreTools:         c.CoreTools,
+		DisableCoreTools:  c.DisableCoreTools,
+		ExtraTools:        c.ExtraTools,
+		ToolWrapper:       c.ToolWrapper,
+		OnMCPServerLoaded: c.OnMCPServerLoaded,
+	}
+	if c.AuthHandler != nil {
+		out.AuthHandler = c.AuthHandler
+	}
+	if c.TokenStoreFactory != nil {
+		out.TokenStoreFactory = tools.TokenStoreFactory(c.TokenStoreFactory)
+	}
+	if c.DebugLogger != nil {
+		out.DebugLogger = c.DebugLogger
+	}
+	out.MCPTaskConfig = c.MCPTaskConfig.toToolsConfig()
+	return out
+}
+
+// ToolCallHandler is invoked when the LLM produces a tool call. It receives
+// the call ID, tool name, and the JSON-encoded input arguments.
+type ToolCallHandler func(toolCallID, toolName, toolArgs string)
+
+// ToolExecutionHandler is invoked at the start and end of tool execution.
+// The isStarting flag distinguishes the two phases.
+type ToolExecutionHandler func(toolCallID, toolName, toolArgs string, isStarting bool)
+
+// ToolResultHandler is invoked after a tool finishes executing. The metadata
+// parameter carries optional structured data (e.g. file-diff info) from the
+// tool execution, JSON-encoded; it may be empty.
+type ToolResultHandler func(toolCallID, toolName, toolArgs, result, metadata string, isError bool)
+
+// ResponseHandler is invoked with the final assistant text for each turn.
+type ResponseHandler func(content string)
+
+// StreamingResponseHandler is invoked with each streamed text delta as it
+// arrives from the LLM.
+type StreamingResponseHandler func(content string)
+
+// ToolCallContentHandler is invoked with any assistant text that accompanies
+// a tool call within the same step.
+type ToolCallContentHandler func(content string)
 
 // ==== Provider & Model Types (internal/models/) ====
 
@@ -126,7 +253,7 @@ type ModelsRegistry = models.ModelsRegistry
 
 // SpinnerFunc wraps a function in a loading spinner animation. Used for
 // Ollama model loading. Signature: func(fn func() error) error.
-type SpinnerFunc = agent.SpinnerFunc
+type SpinnerFunc func(fn func() error) error
 
 // ==== LLM Types ====
 //

pkg/kit/types_test.go

@@ -1,6 +1,7 @@
 package kit_test
 
 import (
+	"context"
 	"encoding/json"
 	"testing"
 
@@ -263,6 +264,101 @@ func TestConvertFromLLMMessage(t *testing.T) {
 	}
 }
 
+// TestAgentConfigNoFantasyImport verifies AgentConfig can be populated with
+// every field — including CoreTools, ExtraTools, and ToolWrapper — using
+// only SDK-owned types. This test deliberately does not import
+// "charm.land/fantasy"; the package compiling at all is the proof that the
+// SDK no longer leaks the dependency name through AgentConfig.
+//
+// Regression test for https://github.com/mark3labs/kit/issues/30.
+func TestAgentConfigNoFantasyImport(t *testing.T) {
+	myTool := kit.NewTool[struct{}]("noop", "does nothing", func(_ context.Context, _ struct{}) (kit.ToolOutput, error) {
+		return kit.TextResult("ok"), nil
+	})
+
+	wrapperCalled := false
+	cfg := kit.AgentConfig{
+		SystemPrompt:     "you are a tester",
+		MaxSteps:         5,
+		StreamingEnabled: true,
+		CoreTools:        []kit.Tool{myTool},
+		ExtraTools:       []kit.Tool{myTool},
+		DisableCoreTools: false,
+		ToolWrapper: func(in []kit.Tool) []kit.Tool {
+			wrapperCalled = true
+			return in
+		},
+		OnMCPServerLoaded: func(_ string, _ int, _ error) {},
+	}
+
+	if cfg.SystemPrompt != "you are a tester" {
+		t.Errorf("SystemPrompt = %q, want %q", cfg.SystemPrompt, "you are a tester")
+	}
+	if cfg.MaxSteps != 5 {
+		t.Errorf("MaxSteps = %d, want 5", cfg.MaxSteps)
+	}
+	if !cfg.StreamingEnabled {
+		t.Error("StreamingEnabled = false, want true")
+	}
+	if len(cfg.CoreTools) != 1 {
+		t.Errorf("CoreTools len = %d, want 1", len(cfg.CoreTools))
+	}
+	if len(cfg.ExtraTools) != 1 {
+		t.Errorf("ExtraTools len = %d, want 1", len(cfg.ExtraTools))
+	}
+
+	// Exercise the wrapper to confirm the func type is usable.
+	out := cfg.ToolWrapper(cfg.CoreTools)
+	if !wrapperCalled {
+		t.Error("ToolWrapper was not invoked")
+	}
+	if len(out) != 1 {
+		t.Errorf("wrapped tool list len = %d, want 1", len(out))
+	}
+}
+
+// TestAgentConfigToolWrapperSignature documents that AgentConfig.ToolWrapper
+// uses kit.Tool (not the underlying provider type) in its signature.
+func TestAgentConfigToolWrapperSignature(t *testing.T) {
+	//nolint:staticcheck // QF1011: explicit type asserts the SDK-side func signature.
+	var _ func([]kit.Tool) []kit.Tool = func(in []kit.Tool) []kit.Tool { return in }
+	cfg := kit.AgentConfig{
+		ToolWrapper: func(in []kit.Tool) []kit.Tool { return in },
+	}
+	if cfg.ToolWrapper == nil {
+		t.Fatal("ToolWrapper assignment failed")
+	}
+}
+
+// TestSpinnerFuncSignature verifies SpinnerFunc has the documented signature
+// and can be constructed without importing any provider package.
+func TestSpinnerFuncSignature(t *testing.T) {
+	called := false
+	var sp kit.SpinnerFunc = func(fn func() error) error {
+		called = true
+		return fn()
+	}
+	err := sp(func() error { return nil })
+	if err != nil {
+		t.Errorf("SpinnerFunc returned err: %v", err)
+	}
+	if !called {
+		t.Error("SpinnerFunc did not invoke fn")
+	}
+}
+
+// TestHandlerTypesSignatures verifies the SDK-owned handler function types
+// can be assigned from plain function literals using only standard library
+// types in their signatures (no provider-package import required).
+func TestHandlerTypesSignatures(t *testing.T) {
+	var _ kit.ToolCallHandler = func(_, _, _ string) {}
+	var _ kit.ToolExecutionHandler = func(_, _, _ string, _ bool) {}
+	var _ kit.ToolResultHandler = func(_, _, _, _, _ string, _ bool) {}
+	var _ kit.ResponseHandler = func(_ string) {}
+	var _ kit.StreamingResponseHandler = func(_ string) {}
+	var _ kit.ToolCallContentHandler = func(_ string) {}
+}
+
 // containsStr is a tiny helper to avoid importing strings in test.
 func containsStr(s, substr string) bool {
 	return len(s) >= len(substr) && (s == substr || len(s) > 0 && indexStr(s, substr) >= 0)

pkg/kit/viper_isolation_test.go

@@ -0,0 +1,232 @@
+package kit_test
+
+import (
+	"context"
+	"os"
+	"testing"
+
+	"github.com/mark3labs/kit/pkg/kit"
+)
+
+// TestOptionFunctionsPlumbing verifies that the functional options apply their
+// values to the underlying Options struct. This does not create a provider, so
+// it runs without API keys.
+func TestOptionFunctionsPlumbing(t *testing.T) {
+	o := &kit.Options{}
+	opts := []kit.Option{
+		kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+		kit.WithSystemPrompt("be terse"),
+		kit.WithMaxTokens(4321),
+		kit.WithThinkingLevel("high"),
+		kit.WithProviderAPIKey("sk-test"),
+		kit.WithProviderURL("https://example.test/v1"),
+		kit.WithConfigFile("/tmp/.kit.yml"),
+		kit.WithStreaming(false),
+		kit.WithDebug(),
+		kit.Ephemeral(),
+	}
+	for _, fn := range opts {
+		fn(o)
+	}
+
+	if o.Model != "anthropic/claude-sonnet-4-5-20250929" {
+		t.Errorf("WithModel: got %q", o.Model)
+	}
+	if o.SystemPrompt != "be terse" {
+		t.Errorf("WithSystemPrompt: got %q", o.SystemPrompt)
+	}
+	if o.MaxTokens != 4321 {
+		t.Errorf("WithMaxTokens: got %d", o.MaxTokens)
+	}
+	if o.ThinkingLevel != "high" {
+		t.Errorf("WithThinkingLevel: got %q", o.ThinkingLevel)
+	}
+	if o.ProviderAPIKey != "sk-test" {
+		t.Errorf("WithProviderAPIKey: got %q", o.ProviderAPIKey)
+	}
+	if o.ProviderURL != "https://example.test/v1" {
+		t.Errorf("WithProviderURL: got %q", o.ProviderURL)
+	}
+	if o.ConfigFile != "/tmp/.kit.yml" {
+		t.Errorf("WithConfigFile: got %q", o.ConfigFile)
+	}
+	if o.Streaming == nil {
+		t.Error("WithStreaming: expected Streaming to be set (non-nil)")
+	} else if *o.Streaming {
+		t.Error("WithStreaming(false): expected *Streaming=false")
+	}
+	if !o.Debug {
+		t.Error("WithDebug: expected Debug=true")
+	}
+	if !o.NoSession {
+		t.Error("Ephemeral: expected NoSession=true")
+	}
+}
+
+// TestOptionOrderingOverrides verifies later options override earlier ones.
+func TestOptionOrderingOverrides(t *testing.T) {
+	o := &kit.Options{}
+	kit.WithModel("a/b")(o)
+	kit.WithModel("c/d")(o)
+	if o.Model != "c/d" {
+		t.Errorf("later WithModel should win; got %q", o.Model)
+	}
+}
+
+// TestKitConfigIsolation is the regression test for issue #40: two Kit
+// instances constructed in the same process must own independent configuration
+// stores. Setting the thinking level (or model) on one must not affect the
+// other. Against the previous global-viper implementation this test fails
+// because both Kits read and write the same process-global store.
+func TestKitConfigIsolation(t *testing.T) {
+	if os.Getenv("ANTHROPIC_API_KEY") == "" {
+		t.Skip("Skipping test: ANTHROPIC_API_KEY not set")
+	}
+
+	ctx := context.Background()
+
+	a, err := kit.New(ctx, &kit.Options{
+		Model:         "anthropic/claude-sonnet-4-5-20250929",
+		ThinkingLevel: "low",
+		Quiet:         true,
+		NoSession:     true,
+		NoExtensions:  true,
+	})
+	if err != nil {
+		t.Fatalf("failed to create Kit A: %v", err)
+	}
+	defer func() { _ = a.Close() }()
+
+	b, err := kit.New(ctx, &kit.Options{
+		Model:         "anthropic/claude-sonnet-4-5-20250929",
+		ThinkingLevel: "high",
+		Quiet:         true,
+		NoSession:     true,
+		NoExtensions:  true,
+	})
+	if err != nil {
+		t.Fatalf("failed to create Kit B: %v", err)
+	}
+	defer func() { _ = b.Close() }()
+
+	// Each instance must retain its own configured thinking level. Under the
+	// old global-viper implementation, B's construction overwrote A's value.
+	if got := a.GetThinkingLevel(); got != "low" {
+		t.Errorf("Kit A thinking level = %q; want %q (config leaked from B)", got, "low")
+	}
+	if got := b.GetThinkingLevel(); got != "high" {
+		t.Errorf("Kit B thinking level = %q; want %q", got, "high")
+	}
+
+	// Mutating one at runtime must not bleed into the other.
+	if err := a.SetThinkingLevel(ctx, "medium"); err != nil {
+		t.Fatalf("SetThinkingLevel on A: %v", err)
+	}
+	if got := a.GetThinkingLevel(); got != "medium" {
+		t.Errorf("after SetThinkingLevel, Kit A = %q; want %q", got, "medium")
+	}
+	if got := b.GetThinkingLevel(); got != "high" {
+		t.Errorf("after mutating A, Kit B leaked to %q; want %q", got, "high")
+	}
+}
+
+// TestNewAgentDefaultsStreamingOn verifies that the ergonomic constructor
+// enables streaming by default and applies functional options.
+func TestNewAgentDefaultsStreamingOn(t *testing.T) {
+	if os.Getenv("ANTHROPIC_API_KEY") == "" {
+		t.Skip("Skipping test: ANTHROPIC_API_KEY not set")
+	}
+
+	ctx := context.Background()
+	k, err := kit.NewAgent(ctx,
+		kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+		kit.WithMaxTokens(2048),
+		kit.Ephemeral(),
+	)
+	if err != nil {
+		t.Fatalf("NewAgent failed: %v", err)
+	}
+	defer func() { _ = k.Close() }()
+
+	if !k.ConfigValueIsSetForTest("max-tokens") {
+		t.Error("NewAgent did not propagate WithMaxTokens to the instance store")
+	}
+	if !k.ConfigBoolForTest("stream") {
+		t.Error("NewAgent should enable streaming by default")
+	}
+}
+
+// TestNewAgentStreamingOptOut verifies WithStreaming(false) disables the
+// default-on streaming behaviour of NewAgent.
+func TestNewAgentStreamingOptOut(t *testing.T) {
+	if os.Getenv("ANTHROPIC_API_KEY") == "" {
+		t.Skip("Skipping test: ANTHROPIC_API_KEY not set")
+	}
+
+	ctx := context.Background()
+	k, err := kit.NewAgent(ctx,
+		kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+		kit.WithStreaming(false),
+		kit.Ephemeral(),
+	)
+	if err != nil {
+		t.Fatalf("NewAgent failed: %v", err)
+	}
+	defer func() { _ = k.Close() }()
+
+	if k.ConfigBoolForTest("stream") {
+		t.Error("WithStreaming(false) should disable streaming")
+	}
+}
+
+// TestNewZeroOptionsKeepsStreamingDefault is the regression test for the
+// unconditional `v.Set("stream", opts.Streaming)` bug: a zero-valued Options
+// (Streaming == nil) must NOT force stream=false. With Streaming unset,
+// streaming resolves through the precedence chain, whose SDK default is true.
+func TestNewZeroOptionsKeepsStreamingDefault(t *testing.T) {
+	if os.Getenv("ANTHROPIC_API_KEY") == "" {
+		t.Skip("Skipping test: ANTHROPIC_API_KEY not set")
+	}
+
+	ctx := context.Background()
+	k, err := kit.New(ctx, &kit.Options{
+		Model:      "anthropic/claude-sonnet-4-5-20250929",
+		Quiet:      true,
+		NoSession:  true,
+		SkipConfig: true, // isolate from any ~/.kit.yml / env stream setting
+	})
+	if err != nil {
+		t.Fatalf("New failed: %v", err)
+	}
+	defer func() { _ = k.Close() }()
+
+	if !k.ConfigBoolForTest("stream") {
+		t.Error("zero-valued Options must not force stream=false; expected the default (true)")
+	}
+}
+
+// TestNewStreamingExplicitOptOut verifies that a raw Options can still disable
+// streaming by setting Streaming to a pointer to false.
+func TestNewStreamingExplicitOptOut(t *testing.T) {
+	if os.Getenv("ANTHROPIC_API_KEY") == "" {
+		t.Skip("Skipping test: ANTHROPIC_API_KEY not set")
+	}
+
+	streamOff := false
+	ctx := context.Background()
+	k, err := kit.New(ctx, &kit.Options{
+		Model:      "anthropic/claude-sonnet-4-5-20250929",
+		Quiet:      true,
+		NoSession:  true,
+		SkipConfig: true,
+		Streaming:  &streamOff,
+	})
+	if err != nil {
+		t.Fatalf("New failed: %v", err)
+	}
+	defer func() { _ = k.Close() }()
+
+	if k.ConfigBoolForTest("stream") {
+		t.Error("Streaming=&false should disable streaming")
+	}
+}

www/pages/configuration.md

@@ -88,6 +88,9 @@ mcpServers:
     type: remote
     url: "https://pubmed.mcp.example.com"
     noOAuth: true  # skip OAuth for public servers
+    headers:
+      - "ApiKey: ${env://API_KEY}"              # required env var
+      - "X-Tenant: ${env://TENANT_ID:-default}" # with fallback default
 
   builds:
     type: remote
@@ -106,9 +109,10 @@ mcpServers:
 | `allowedTools` | list | Whitelist of tool names to expose |
 | `excludedTools` | list | Blacklist of tool names to hide |
 | `noOAuth` | bool | Skip OAuth for this server (for public servers that don't require auth) |
+| `headers` | list of strings | HTTP headers to attach to every request, each as a `"Key: Value"` string. Values support env-substitution: `${env://VAR}` or `${env://VAR:-default}`. |
 | `tasksMode` | string | When to augment `tools/call` with MCP task metadata: `auto` (default — only when the server advertises task support), `never`, or `always`. See [MCP tasks](#mcp-tasks-long-running-tools). |
 
-A legacy format with `transport`, `args`, `env`, and `headers` fields is also supported.
+A legacy format with `transport`, `args`, and `env` fields is also supported; `headers` works in both the current and legacy formats.
 
 ### MCP tasks (long-running tools)
 

www/pages/providers.md

@@ -143,6 +143,39 @@ The `custom/custom` model has zero cost, 262K context window, and supports reaso
 
 Optionally set `CUSTOM_API_KEY` environment variable or use `--provider-api-key` for endpoints requiring authentication.
 
+## Auto-routed providers
+
+Any provider in the [models.dev](https://models.dev) database can be used with the
+standard `provider/model` format, even without a dedicated native integration. Kit
+auto-routes the request through the matching **wire protocol** — the actual API
+shape the provider speaks — rather than requiring a per-provider code path:
+
+| Wire protocol | npm package (models.dev) | Transport used |
+|---------------|--------------------------|----------------|
+| OpenAI (Responses API) | `@ai-sdk/openai` | OpenAI |
+| OpenAI (chat completions) | `@ai-sdk/openai-compatible` | OpenAI-compatible |
+| Anthropic | `@ai-sdk/anthropic` | Anthropic |
+| Google Gemini | `@ai-sdk/google` | Google |
+
+The provider's `api` URL from the database is used as the base URL. A provider
+whose npm package isn't recognized but that has an `api` URL falls back to the
+OpenAI-compatible wire.
+
+Because routing follows the wire protocol, aggregator/proxy providers work across
+**all** of their models — including ones they re-flavor onto a different protocol
+via a per-model override. For example, an aggregator that proxies Claude, GPT,
+*and* Gemini routes them to the Anthropic, OpenAI, and Google transports
+respectively:
+
+```bash
+kit --model opencode/claude-haiku-4-5 "Hello"     # → Anthropic wire
+kit --model opencode/gpt-5 "Hello"                # → OpenAI wire
+kit --model opencode/gemini-3.5-flash "Hello"     # → Google wire
+```
+
+Provide the provider's API key the same way as any other — via its environment
+variable (e.g. `OPENCODE_API_KEY`) or `--provider-api-key`.
+
 ## Model database
 
 Kit ships with a local model database that maps provider names to API configurations. You can manage it with:

www/pages/sdk/options.md

@@ -7,6 +7,16 @@ description: Configuration options for the Kit Go SDK.
 
 Pass an `Options` struct to `kit.New()` to configure the Kit instance.
 
+::: tip
+For simple setups, `kit.NewAgent(ctx, ...Option)` provides functional-options
+helpers (`WithModel`, `WithStreaming`, `Ephemeral`, ...) over the same `Options`
+struct. See [Functional options](/sdk/overview#functional-options-newagent).
+:::
+
+Each `kit.New` / `kit.NewAgent` call owns an isolated configuration store, so
+these options never leak between Kit instances in the same process. See
+[Per-instance config isolation](/sdk/overview#per-instance-config-isolation).
+
 ## Full options reference
 
 ```go
@@ -18,7 +28,7 @@ host, err := kit.New(ctx, &kit.Options{
 
     // Behavior
     MaxSteps:     10,
-    Streaming:    true,
+    Streaming:    ptrBool(true), // *bool: nil = unset (default true), &false = off
     Quiet:        true,
     Debug:        true,
 
@@ -91,7 +101,7 @@ host, err := kit.New(ctx, &kit.Options{
 | `SystemPrompt` | `string` | — | System prompt text or file path |
 | `ConfigFile` | `string` | `~/.kit.yml` | Path to config file |
 | `MaxSteps` | `int` | `0` | Max agent steps (0 = unlimited) |
-| `Streaming` | `bool` | `true` | Enable streaming output |
+| `Streaming` | `*bool` | `nil` | Enable streaming output. `nil` leaves it to the precedence chain (env → config → default `true`); `&true`/`&false` forces it. Pointer so unset is distinct from explicit `false`. |
 | `Quiet` | `bool` | `false` | Suppress output |
 | `Debug` | `bool` | `false` | Enable debug logging |
 
@@ -114,9 +124,10 @@ defaults for samplers).
 | `FrequencyPenalty` | `*float32` | — | OpenAI-family frequency penalty. `nil` leaves provider default. |
 | `PresencePenalty` | `*float32` | — | OpenAI-family presence penalty. `nil` leaves provider default. |
 
-Pointer-typed samplers are populated via a tiny helper:
+Pointer-typed fields (`Streaming` and the samplers) are populated via tiny helpers:
 
 ```go
+func ptrBool(v bool) *bool          { return &v }
 func ptrFloat32(v float32) *float32 { return &v }
 ```
 
@@ -127,7 +138,7 @@ when embedding Kit as a library.
 
 | Field | Type | Default | Description |
 |-------|------|---------|-------------|
-| `ProviderAPIKey` | `string` | — | API key used to authenticate with the provider. `""` falls back to config / provider-specific env var (e.g. `ANTHROPIC_API_KEY`). When set, overrides any pre-existing viper state. |
+| `ProviderAPIKey` | `string` | — | API key used to authenticate with the provider. `""` falls back to config / provider-specific env var (e.g. `ANTHROPIC_API_KEY`). When set, it takes precedence over config and env values on this instance's store. |
 | `ProviderURL` | `string` | — | Override the provider endpoint (e.g. LiteLLM, vLLM, Azure OpenAI, internal proxy). `""` = provider default. |
 | `TLSSkipVerify` | `bool` | `false` | Disable TLS certificate verification on the provider HTTP client. Only effective when `true`; to force-disable, use config file or env var instead. For self-signed dev certs only. |
 
@@ -160,6 +171,14 @@ when embedding Kit as a library.
 | `SkillsDir` | `string` | — | Override default skills directory |
 | `NoSkills` | `bool` | `false` | Disable skill loading entirely |
 
+These fields only control the **initial** skill and context-file set picked
+up by `New()`. To add, remove, or replace skills and `AGENTS.md`-style
+context files at runtime (e.g. per user or per session), use the
+`AddSkill` / `LoadAndAddSkill` / `RemoveSkill` / `SetSkills` and
+`AddContextFile` / `AddContextFileContent` / `RemoveContextFile` /
+`SetContextFiles` methods on `*kit.Kit`. See
+[Runtime skills and context files](/sdk/overview#runtime-skills-and-context-files).
+
 ### Compaction & MCP
 
 | Field | Type | Default | Description |

www/pages/sdk/overview.md

@@ -45,6 +45,73 @@ func main() {
 }
 ```
 
+## Functional options (`NewAgent`)
+
+For simple programmatic setups, `kit.NewAgent` offers an ergonomic
+functional-options front door over `kit.New`. Streaming is **enabled by
+default**; pass `kit.WithStreaming(false)` to opt out.
+
+```go
+host, err := kit.NewAgent(ctx,
+    kit.WithModel("anthropic/claude-sonnet-4-5-20250929"),
+    kit.WithSystemPrompt("You are a helpful assistant."),
+    kit.WithMaxTokens(8192),
+    kit.WithThinkingLevel("medium"),
+    kit.Ephemeral(), // in-memory session, no persistence
+)
+if err != nil {
+    log.Fatal(err)
+}
+defer host.Close()
+```
+
+Available options:
+
+| Option | Sets |
+|--------|------|
+| `WithModel(string)` | `Options.Model` (provider/model format) |
+| `WithSystemPrompt(string)` | `Options.SystemPrompt` (inline text or file path) |
+| `WithStreaming(bool)` | `Options.Streaming` (default `true` under `NewAgent`) |
+| `WithMaxTokens(int)` | `Options.MaxTokens` |
+| `WithThinkingLevel(string)` | `Options.ThinkingLevel` |
+| `WithTools(...Tool)` | `Options.Tools` (replaces the default set) |
+| `WithExtraTools(...Tool)` | `Options.ExtraTools` (adds alongside defaults) |
+| `WithProviderAPIKey(string)` | `Options.ProviderAPIKey` |
+| `WithProviderURL(string)` | `Options.ProviderURL` |
+| `WithConfigFile(string)` | `Options.ConfigFile` |
+| `WithDebug()` | `Options.Debug = true` |
+| `Ephemeral()` | `Options.NoSession = true` |
+
+Options are applied in order, so later options override earlier ones. `Option`
+is a plain `func(*Options)`, so you can define your own. For advanced
+configuration not covered by the helpers (custom MCP config, in-process MCP
+servers, session backends, MCP task tuning) construct an `Options` value
+explicitly and call `kit.New`.
+
+### When to use which
+
+| Constructor | Use when |
+|-------------|----------|
+| `kit.NewAgent(ctx, ...Option)` | Quick programmatic setups; you only need the common fields. Streaming defaults on. |
+| `kit.New(ctx, *Options)` | You need fields without a `With*` helper (`MCPConfig`, `InProcessMCPServers`, `SessionManager`, MCP task tuning, etc.), or you already hold an `Options` value. |
+
+## Per-instance config isolation
+
+Each `kit.New` / `kit.NewAgent` call owns an **isolated configuration store**,
+so constructing multiple Kit instances in the same process is safe: setting the
+model, thinking level, or generation parameters on one never affects another,
+and runtime mutators (`SetModel`, `SetThinkingLevel`) only touch the owning
+instance. This makes subagent spawning and multi-Kit embedding race-free with
+no external synchronization required.
+
+```go
+a, _ := kit.NewAgent(ctx, kit.WithThinkingLevel("low"))
+b, _ := kit.NewAgent(ctx, kit.WithThinkingLevel("high"))
+
+a.SetThinkingLevel(ctx, "medium")
+// a.GetThinkingLevel() == "medium"; b.GetThinkingLevel() is still "high"
+```
+
 ## Multi-turn conversations
 
 Conversations retain context automatically across calls:
@@ -201,6 +268,66 @@ host, _ := kit.New(ctx, &kit.Options{
 n, _ := host.AddInProcessMCPServer(ctx, "docs", mcpSrv)
 ```
 
+## Runtime skills and context files
+
+Kit auto-discovers skills and `AGENTS.md`-style context files during `New()`,
+but multi-tenant hosts (chatbots, web services, per-user agents) often need
+to swap these **after** construction. The runtime mutators below recompose
+the system prompt and apply it to the agent so the next turn picks up the
+updated instructions — no restart, no file shuffling.
+
+```go
+// Add a programmatic skill — no file on disk required.
+host.AddSkill(&kit.Skill{
+    Name:        "polite-french",
+    Description: "Respond in French and always greet the user.",
+    Content:     "Always reply in French. Open every response with 'Bonjour'.",
+})
+
+// Or load one from disk.
+host.LoadAndAddSkill("/var/skills/refund-policy.md")
+
+// Project context (AGENTS.md equivalents): inline content from a DB...
+host.AddContextFileContent(
+    fmt.Sprintf("session://%s/AGENTS.md", userID),
+    rulesFromDB,
+)
+// ...or load from disk.
+host.LoadAndAddContextFile("/etc/agents/tenant-acme.md")
+
+// Remove individually when a session ends.
+host.RemoveSkill("polite-french")
+host.RemoveContextFile(fmt.Sprintf("session://%s/AGENTS.md", userID))
+
+// Or replace the whole set in one call.
+host.SetSkills(activeSkillsForUser)
+host.SetContextFiles(activeContextForUser)
+
+// Inspect current state (snapshot copies — safe to mutate).
+skills := host.GetSkills()
+ctxFiles := host.GetContextFiles()
+```
+
+Key points:
+
+- **Auto-refresh.** Every `Add*` / `Remove*` / `Set*` call recomposes the system
+  prompt against the captured base prompt (preserving per-model overrides and
+  `--system-prompt` resolution) and pushes the result onto the agent. Call
+  `host.RefreshSystemPrompt()` only if you mutate state through a different
+  path and need to force a re-render.
+- **Dedup keys.** Skills dedupe by `Name`; context files dedupe by `Path`.
+  Re-adding the same key replaces the entry instead of appending a duplicate.
+- **Path is opaque.** `ContextFile.Path` does not have to point at a real file
+  — it's only used for dedup and for the `Instructions from: <Path>` header
+  injected into the prompt. URIs like `session://user-123/AGENTS.md` work fine.
+- **Thread safety.** All readers and mutators are safe to call concurrently
+  from multiple goroutines; the underlying state is guarded by an internal
+  `RWMutex`.
+- **Init-time options still apply.** `Options.Skills`, `Options.SkillsDir`,
+  `Options.NoSkills`, and `Options.NoContextFiles` continue to control the
+  startup set; the runtime API mutates from whatever state `New()` produced.
+  See [SDK options](/sdk/options#skills--configuration).
+
 ## MCP prompts and resources
 
 Query prompts and resources exposed by connected MCP servers: