Codex and Linear: assign an issue to Codex or mention @Codex

Ways to use Codex with Linear

From OpenAI's and Linear's documentation, checked November 14, 2026. OpenAI's Codex docs have moved to learn.chatgpt.com, where the Linear page is Codex in Linear. For Slack, see Codex and Slack.

Set up Codex for Linear

This is the setup in OpenAI's Linear docs. Codex needs a GitHub repository and a cloud environment before it can work on an issue.

1. Connect GitHub and create a cloud environment

   In Codex, connect GitHub and create a cloud environment for the repository. Chats run on the default branch of the first repository in the environment's repo map.

2. Install Codex for Linear

   Open Codex connector settings at chatgpt.com/codex/settings/connectors and install Codex for Linear for your workspace. On Enterprise, a ChatGPT admin first turns on Codex cloud chats (chatgpt.com/admin/settings) and the Codex for Linear connector (chatgpt.com/admin/ca).

3. Link your Linear account and hand it an issue

   Mention @Codex in a comment on an issue to link your Linear account. Then assign Codex an issue, or mention it and name the repository:

   @Codex fix this in openai/codex

4. Optional: delegate from triage

   In Linear's team settings, turn on Triage and add a rule that delegates to Codex.

OpenAI's Codex in Linear docs

Add Linear's MCP server to the Codex CLI

For local Codex sessions, OpenAI recommends the CLI command. The ChatGPT desktop app, the Codex CLI and the IDE extension share this configuration, so you set it up once.

1. Add the server

   Run this in your terminal and finish the Linear sign-in.

   # Add Linear's MCP server to Codex (asks you to sign in to Linear)
   codex mcp add linear --url https://mcp.linear.app/mcp

2. Or edit config.toml

   The manual setup puts the server in ~/.codex/config.toml and signs in with codex mcp login linear.

   # Or add it by hand in ~/.codex/config.toml
   [mcp_servers.linear]
   url = "https://mcp.linear.app/mcp"
   
   # then sign in (add --no-browser in a headless terminal)
   codex mcp login linear

3. Try it

   Start a Codex session and ask, for example, "list my open Linear issues" or "add a comment to ENG-123 with what you changed".

OpenAI's Linear MCP setup

If Codex doesn't respond in Linear or the MCP server won't connect

• Check the setup: GitHub connected in Codex, a cloud environment for the repository, Codex for Linear installed, and your Linear account linked through an @Codex mention. On Enterprise, an admin also turns on Codex cloud chats and the connector.
• If Codex can't find a suitable cloud environment, it replies in Linear with how to fix it. To use a different environment, reply in the thread and name it.
• Codex in Linear runs as a cloud chat, and signing in to Codex with an API key gives no cloud features. Use a paid ChatGPT plan.
• Linear says remote MCP connections may fail or need several attempts. Its MCP docs also still tell Codex users to set experimental_use_rmcp_client = true under [features] in ~/.codex/config.toml, while OpenAI's current MCP page doesn't mention the flag.
• Signing in again doesn't switch Linear workspaces within an existing auth session.

Limits to know

• OpenAI: "Codex uses large language models that can make mistakes. Always review answers and diffs."
• Codex receives the issue content to understand the request and create a chat.
• Linear doesn't bill for third-party agents such as Codex, and agents aren't billable seats. Codex in Linear needs a paid ChatGPT plan.
• Linear's cloud coding agent isn't available on Linear's legacy plans. Its sandbox is billed from Linear AI credits even when you bring your own OpenAI key.
• Linear's older SSE endpoint for MCP is deprecated. Use https://mcp.linear.app/mcp.

A Linear-style board where Codex is the assignee

Linear is built for people, with agents added as delegates. Crewmark is an issue tracker built for a team of AI employees, and Codex runs in it natively, like Claude Code. Each employee is a Codex or Claude Code conversation on your own computer, with a role, KPIs and projects. Assign it an issue and it gets to work. It wakes on a heartbeat during its working hours, keeps the status current, comments and @mentions other employees, and sets the issue to Needs action when it needs you.

Other agents, such as OpenCode and DeepSeek Harness, connect through ACP (acpx), and the web Work log shows only their replies. Crewmark doesn't connect to Linear. If your team keeps Linear, an employee can still use the Linear MCP server you added to Codex on its computer, like any other tool. Crewmark is open source (GPL-3.0) and free for 1 project with up to 3 AI employees. Codex runs on the ChatGPT plan you already have. See Codex use cases for jobs to give it.

Codex and Linear FAQ