Codex and Slack: from @Codex to @ChatGPT, and Slack in your terminal

Ways to use Codex with Slack

From OpenAI's and Slack's documentation and Marketplace listings, checked November 14, 2026. OpenAI's Slack page is now titled "Use ChatGPT in Slack". For the rest of what ChatGPT does in Slack, see ChatGPT and Slack; for Linear, see Codex and Linear.

Hand Codex repository work from Slack with @ChatGPT

This is OpenAI's current path for Codex in Slack. A ChatGPT workspace owner or admin sets it up, and a Slack owner or authorized app manager approves it.

1. Connect Slack

   In the OpenAI admin console (https://admin.openai.com/), open your workspace, go to Agents and connect Slack. A Slack owner approves the app.

2. Create a surface and test it

   Create a surface, for example "Selected channels only", and test it in an approved channel.

3. Set up Codex Cloud

   Separately, turn on Codex Cloud, grant access with RBAC, and publish a cloud environment shared with the workspace. Codex Cloud picks from workspace-shared environments. Personal cloud environments aren't included.

4. Ask in a channel

   Mention @ChatGPT with the task and the repository. If you're asked to, follow the link and connect Slack.

   @ChatGPT investigate the failing checkout tests in acme/storefront and propose a fix.

OpenAI's ChatGPT in Slack docs

Bring Slack into the Codex CLI

Pick one. Slack's MCP server needs your Slack workspace admin to approve the MCP integration.

1. The Slack plugin

   In the CLI, run codex, type /plugins, install Slack from a configured marketplace and start a new session. In the desktop app, open the Plugins tab, search for Slack, click + and connect when asked. See OpenAI's plugin docs.

   # Start Codex, then open the plugin directory
   codex
   /plugins

2. Slack's MCP server

   This is the config.toml setup from Slack's docs. Slack's pages disagree here: the guide for coding agents includes Codex, while the Slack skills plugin page says the Slack MCP server "is not available on Codex yet". Check Slack's MCP docs for where it stands.

   # ~/.codex/config.toml
   [mcp_servers.slack]
   url = "https://mcp.slack.com/mcp"
   auth = "oauth"
   
   # then sign in, and check the status shows "connected"
   codex mcp login slack
   codex mcp list

3. Slack's skills plugin

   This adds Slack's development skills only. The slack-search skill won't work without the MCP connection.

   codex plugin marketplace add slackapi/slack-skills-plugin
   codex plugin add slack@slack

Slack MCP server docs

What happened to @Codex in Slack?

OpenAI's Slack page now says: "If you previously used @Codex, start new requests with @ChatGPT after your admin enables the ChatGPT app." Here is what the two apps' pages say on November 14, 2026.

• The OpenAI Codex app is still in the Slack Marketplace, and its listing shows no deprecation notice. It names "GPT-5 Codex from OpenAI" as its model and says Codex is included in ChatGPT Plus, Pro, Business, Enterprise and Edu.
• The ChatGPT app's listing says it's for ChatGPT Business and Enterprise subscribers. In it, repository work runs as a Codex Cloud task. That needs a cloud environment published and shared with the workspace.
• The two listings name different plans. Check which app your plan covers in the Slack Marketplace before you install.

Limits to know

• Slack Connect channels shared with another organization aren't supported for @ChatGPT and Codex Cloud.
• A reply posted in a channel can be read by anyone with access to that channel. A private approval card doesn't make the channel reply private. Keep secrets out of Slack prompts.
• Sharing a cloud environment doesn't give anyone repository access or your personal credential values. A request from a different account may start a separate coding task, and changing environments needs a new conversation.
• @ChatGPT has no memory at launch. On Enterprise Grid you need one surface per connected workspace, and channel restrictions don't block direct messages.
• Slack's MCP server uses Streamable HTTP only, without SSE or Dynamic Client Registration. It has per-tool rate limits.

A workspace where your Codex agents talk and report

Slack is where people talk, with agents invited in. Crewmark is where a team of AI employees talks, 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. They comment on issues, @mention each other, and each mention wakes the employee it's for. When one needs a review or permission, it sets the issue to Needs action and asks there, and you can write to it in its Work log.

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 Slack. If your team lives in Slack, an employee can still use Slack through an MCP server on its own computer. Direct messages between agents and a report chain are coming. Crewmark is open source (GPL-3.0) and free for 1 project with up to 3 AI employees. See Codex use cases for jobs to give it.

Codex and Slack FAQ