lobe-chat/docs/usage/agent/codex.mdx

---
title: Codex
description: >-
  Delegate OpenAI Codex inside LobeHub — chat with the Codex CLI from your
  desktop app, watch file changes, todos, and command output stream in real
  time, and resume sessions across turns.
tags:
  - LobeHub
  - Codex
  - Coding Agent
  - Desktop
  - CLI
  - OpenAI
---

# Codex

Codex is OpenAI's coding agent that edits files, runs commands, and ships changes from your terminal. In LobeHub, you can delegate Codex from the desktop app — keep the chat UX you already use, while Codex does the work locally with full access to your project.

Send a prompt and Codex opens files, makes edits, runs tests, and reports back. File changes, todos, and command output stream into the chat as the agent moves; sessions resume across turns so a long task can span many messages.

## What Is Codex in LobeHub?

A bridge between LobeHub's chat UI and the Codex CLI running on your machine. LobeHub spawns the Codex CLI as a local subprocess, streams its events into a chat conversation, and renders Codex's tool output — file changes, todo lists, command runs — as first-class chat blocks. You drive the agent in natural language; Codex executes locally with your environment, credentials, and project context.

## Requirements

- **LobeHub desktop app** — Codex agents only work in the desktop build. The web app cannot spawn local processes.
- **Codex CLI installed** — the `codex` command must be available on your `PATH`.
- **Signed in** — you must run `codex` once in a terminal to authenticate before LobeHub can drive it.

## Install the Codex CLI

Pick one of the install paths:

**Recommended (npm)**

```bash
npm install -g @openai/codex
```

**Homebrew (macOS)**

```bash
brew install --cask codex
```

After installing, run `codex` once in a terminal to sign in. See the [Codex installation guide](https://github.com/openai/codex#installing-and-running-codex-cli) for details.

If LobeHub can't find the CLI, it shows an **Install Codex CLI** prompt with the same commands and an **Open System Tools** button — click it after installing to re-detect the CLI.

## Add Codex in LobeHub

When LobeHub detects the Codex CLI on your machine, an **Add Codex** recommendation card appears on the home page tagged "Coding Agent". Click it to create a Codex agent in one step.

You can also create one manually from the **Create Agent** menu and pick **Codex** as the type.

Each agent is independent, so you can keep multiple Codex agents pinned to different projects or workflows.

## Working Directory

Every Codex session is pinned to a working directory — the folder Codex sees as the project root. Set it from the chat input bar before sending your first message. Switching the working directory mid-conversation starts a new Codex session for the topic; chat history stays, but the previous session context cannot be resumed.

If you change folders and the saved Codex thread can't be resumed safely, LobeHub shows: *"The saved Codex thread could not be resumed safely, so a new conversation has started for this topic."*

## What Gets Rendered in Chat

LobeHub renders Codex's tool calls with purpose-built blocks instead of raw JSON:

**File changes** — Codex's edits show up as an expandable list with the operation kind (added, deleted, modified, renamed), the file path, and a per-file line count delta (+/−). Click to see what changed.

**Todo lists** — When Codex plans a multi-step task, the plan renders as a progress card with completed / in-progress / pending items and a running count (e.g. "3/5 completed"). Watch tasks tick off as Codex finishes them.

**Command execution** — Shell commands Codex runs show the command, exit code, and stdout / stderr output. Success and failure states are clearly marked.

**Subagents** — Codex can spawn subagents to work in parallel. Their work appears in isolated threads inside the conversation without leaking into the main bubble.

## Sessions and Resume

Codex sessions persist across messages in the same topic. You can send a follow-up like "now also update the tests" and Codex picks up where it left off — same files, same context, same plan.

A session can't be resumed if:

- The working directory changed since the saved thread was created
- The original Codex thread no longer exists
- The CLI returns a "no conversation found" or "thread not found" error

In any of these cases, LobeHub starts a fresh conversation automatically.

## Where It Can Run

The **Execution Device** selector lets you pick where the Codex agent runs:

- **This device** — runs Codex as a local process inside the desktop app. Default.
- **Cloud sandbox** — runs Codex in an ephemeral cloud sandbox. Useful when you don't want the agent touching your local filesystem.
- **Remote device** — drives a remote machine you've connected with `lh connect`. Useful when the project lives on a different machine.

## Limitations

- **Desktop only** — the Codex agent runs in the LobeHub desktop app. The web app cannot spawn the CLI.
- **One sign-in per machine** — Codex shares its authentication with the global CLI. If `codex` works in your terminal, it works in LobeHub.
- **Working-directory-bound** — sessions don't follow you across folders or machines.

## Tips

- **Run `codex` once in a terminal first** — sign-in happens at the CLI level, not in LobeHub.
- **Pick the working directory before your first message** — switching it later starts a new session.
- **Watch the todo card** — it's the fastest read on what Codex thinks it still has to do.
- **Use one Codex agent per project** — pinning each agent to a specific repo keeps sessions tidy and resumable.

<Cards>
  <Card href={'/docs/usage/agent/claude-code'} title={'Claude Code'} />

  <Card href={'/docs/usage/agent/agent-team'} title={'Agent Groups'} />

  <Card href={'/docs/usage/agent/sandbox'} title={'Cloud Sandbox'} />
</Cards>