Codex Plugin for Claude Code: Setup and Commands
The Codex plugin for Claude Code, from OpenAI: install it, run Codex reviews and hand off tasks from Claude Code, the review gate, and what it costs to use.
Josh Pocock7 min read
AI Coding
Codex in Claude Code
Stride AI Academy
The Codex plugin for Claude Code is an open-source plugin from OpenAI that lets you run Codex from inside a Claude Code session. You install it from OpenAI's GitHub repository with two /plugin commands, then use slash commands like /codex:review to get a Codex code review or /codex:rescue to hand a task to Codex. It runs through the Codex CLI on your machine, so it uses your existing ChatGPT or OpenAI API sign-in and counts toward your Codex usage limits.
This guide covers what the plugin does, how to install it, every command, the optional review gate, what it costs and when it's worth using. Everything here was checked against the plugin's repository at github.com/openai/codex-plugin-cc (published in OpenAI's GitHub organization, Apache-2.0 licensed), Anthropic's plugin install docs and Codex pricing on November 22, 2026. Stride AI Academy teaches both Claude Code and Codex and isn't affiliated with Anthropic or OpenAI.
What the Codex plugin does
The repository describes it in one line: use Codex from Claude Code to review code or delegate tasks. In practice that gives you three jobs:
- A second opinion on your code. Codex reviews your uncommitted changes or your branch against
main, without editing anything. - A tougher review. An adversarial mode questions design choices, hidden assumptions and failure modes.
- A handoff. You give Codex a task, such as tracking down a bug or trying a fix, and it works on it, in the background if you like.
It isn't a separate runtime. The README says it delegates through your local Codex CLI and Codex app server, using the same install, sign-in, config and repository checkout. If Codex already works in your terminal, the plugin uses that setup.
If you haven't used Codex on its own yet, our Codex CLI tutorial covers it, and Claude Code vs Codex compares the two agents side by side.
What you need
From the README:
- A ChatGPT subscription, including Free, or an OpenAI API key. Usage counts toward your Codex usage limits.
- Node.js 18.18 or later.
- Claude Code, with plugin support. See how to install Claude Code if you haven't.
How to install the Codex plugin
Run these inside a Claude Code session, in order.
1. Add OpenAI's marketplace:
/plugin marketplace add openai/codex-plugin-cc2. Install the plugin:
/plugin install codex@openai-codexAnthropic's plugin docs say this opens the plugin's details in the /plugin panel first, so you can see what it adds and pick a scope: just you in every project, everyone on this repository, or just you in this repository.
3. Reload plugins:
/reload-plugins4. Check Codex is ready:
/codex:setup/codex:setup checks that Codex is installed and signed in. The README says it can offer to install Codex with npm if it's missing. To install Codex yourself instead:
npm install -g @openai/codexIf Codex isn't signed in, run this from inside Claude Code (the ! runs it as a shell command):
!codex logincodex login supports signing in with a ChatGPT account or an API key.
Before installing any plugin from a marketplace you don't know, Anthropic's docs suggest reading the details pane, since a plugin can run hooks and MCP servers. Their docs also note that auto-update is off by default for third-party marketplaces, so update this one yourself from the Marketplaces tab in /plugin. For how plugins work in general, see Claude Code plugins.
Codex plugin commands
| Command | What it does |
|---|---|
/codex:review |
Read-only Codex review of uncommitted changes, or your branch against a base with --base |
/codex:adversarial-review |
Read-only review that challenges design choices and tradeoffs. Accepts focus text |
/codex:rescue |
Hands a task to Codex through a codex:codex-rescue subagent |
/codex:transfer |
Creates a Codex thread from your current Claude Code session |
/codex:status |
Shows running and recent jobs for this repository |
/codex:result |
Shows the final output of a finished job |
/codex:cancel |
Cancels a background job |
/codex:setup |
Checks install and sign-in, and turns the review gate on or off |
/codex:review
The standard review. Run it before you commit or open a pull request:
/codex:review --base mainIt accepts --wait and --background, but not custom focus text. The README says reviews of multi-file changes can be slow, so it recommends running them in the background.
/codex:adversarial-review
Use this when you want Codex to push back. It's steerable, so you can tell it where to look:
/codex:adversarial-review check the retry logic for race conditionsIt reviews only. It doesn't fix code.
/codex:rescue
Hands work to Codex, such as investigating a bug, attempting a fix or continuing an earlier task. It takes --background, --wait, --resume, --fresh, --model and --effort. If you leave out model and effort, Codex uses its own defaults.
/codex:rescue --background find out why the login test fails on CIBackground jobs
Long reviews and tasks are best started with --background. Then:
/codex:statusto see what's running/codex:resultto read the output, including the Codex session ID when there is one/codex:cancelto stop a job
To carry a job on in Codex itself, the README says to run codex resume <session-id> with the ID from /codex:result or /codex:status.
The review gate
The plugin has an optional review gate. Turn it on with:
/codex:setup --enable-review-gateWhen it's on, a Stop hook runs a targeted Codex review of Claude's response each time Claude tries to finish. If Codex finds issues, Claude's stop is blocked so it can fix them first. Turn it off with --disable-review-gate.
The README carries a clear warning: this can create a long-running loop between Claude and Codex and drain usage limits quickly. Only enable it when you're watching the session. If hooks are new to you, our Claude Code hooks tutorial explains how a Stop hook works.
Config: the plugin uses your Codex settings
The plugin reads your normal Codex config. Set defaults such as model and model_reasoning_effort in ~/.codex/config.toml for your user, or .codex/config.toml in a project. The README says project-level settings only load when the project is trusted.
model_reasoning_effort = "high"An existing API key or base URL setting still applies. Our Codex config.toml guide covers the file in full.
What it costs
The plugin itself is free and open source. What you pay for is Codex usage, which counts against your Codex plan or bills to your API key. From Codex's pricing page, November 22, 2026:
| Codex plan | Price |
|---|---|
| Free | $0/month |
| Go | $8/month |
| Plus | $20/month |
| Pro | From $100/month |
| Business | $20/user/month billed annually, or $25 billed monthly |
| API key | Pay per use at API rates |
Your Claude Code usage still counts against your Claude plan as normal. So using the plugin means drawing on two allowances at once: Claude for the main session, Codex for reviews and handoffs. For Codex's limits in detail, see how much Codex costs, and for Claude Code's side, how much Claude Code costs.
When the Codex plugin is worth using
- Reviews before you ship. A second agent from another company reads your diff with fresh eyes. Run
/codex:reviewbefore each pull request. - When Claude is stuck. Hand the bug to
/codex:rescuein the background and keep working. - Design checks.
/codex:adversarial-reviewwith a focus line is a quick way to stress-test an approach before you build on it.
Skip it if your Codex limits are already tight, since reviews of large diffs use them up, or if you don't want a second company's model reading your client's code. Check the client agrees before you send their code to another provider.
For Anthropic's own review tooling, see Claude Code review.
How to remove it
Open /plugin, go to the Installed tab, select the Codex plugin and choose Uninstall. From your shell:
claude plugin uninstall codex@openai-codexLearn to build with Claude Code
The Stride AI Coding course teaches Claude Code and Codex for building client tools, written for non-programmers, with weekly live calls to review your builds. Many of the free guides cover Claude Code setups, and the Claude Code community is where members compare theirs.
FAQ
Is there an official Codex plugin for Claude Code?
Yes. OpenAI publishes codex-plugin-cc in its own GitHub organization at github.com/openai/codex-plugin-cc, under the Apache-2.0 license. It lets you run Codex reviews and delegate tasks from inside Claude Code.
How do I install the Codex plugin in Claude Code?
In a Claude Code session, run /plugin marketplace add openai/codex-plugin-cc, then /plugin install codex@openai-codex, then /reload-plugins. Finish with /codex:setup to check Codex is installed and signed in.
Do I need a paid ChatGPT plan for the Codex plugin?
No. The README says a ChatGPT subscription including Free, or an OpenAI API key, works. Usage counts toward your Codex limits, so heavy use fits a paid plan better.
Does the Codex plugin use my Claude usage?
Codex's work counts against your Codex plan or API key. Your Claude Code session, including the turns where Claude starts a Codex job and reads the result, still counts against your Claude plan.
What does /codex:rescue do?
It hands a task to Codex, such as investigating a bug or attempting a fix. You can run it in the background, resume an earlier task, and set the model and effort. Check on it with /codex:status and read the output with /codex:result.
What is the Codex review gate?
An optional Stop hook that runs a Codex review each time Claude tries to finish, and blocks the stop if Codex finds issues. The README warns it can loop and drain usage quickly, so only turn it on when you're watching.