Codex plugin for Claude Code: handing repo tasks to a second AI without leaving the terminal
This plugin lets Claude Code send repository tasks to Codex for implementation or a second code review.
At a glance
- What is it?
- The openai/codex-plugin-cc marketplace plugin adds Codex review and delegation commands to Claude Code. It is a pragmatic bridge for teams already on Claude Code who want Codex's different model behavior on the same repository.
- Who is it for?
- Adopt this plugin if you are a Claude Code user who wants a second opinion from Codex on the same repo, especially for read-only reviews or background delegation. Skip it if you have no ChatGPT subscription or OpenAI API key, if you are on Node.js below 18.18, or if you need a steerable review that also fixes code; /codex:adversarial-review is read-only and /codex:rescue is the only path to changes.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 70 days ago.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What this plugin actually solves
The plugin solves a workflow gap: Claude Code users who want Codex's judgment on the same repository, without switching terminals or copying diffs. It adds slash commands that send the current uncommitted changes or a branch diff to Codex for review, and it can delegate implementation tasks to a Codex subagent. The intended user is someone already comfortable with Claude Code who wants a second AI model to look at the same codebase. The README is explicit that the review quality matches running /review inside Codex directly, so this is not a new review engine, it is a transport layer between two tools.
The command surface: review, rescue, transfer, and job management
The plugin exposes eight slash commands. /codex:review runs a normal, read-only Codex review on current work or a branch with --base <ref>. /codex:adversarial-review is the same target selection but adds steerability: you can pass focus text after the flags, like 'challenge whether this was the right caching and retry design'. It still does not fix code. /codex:rescue hands a task to the codex:codex-rescue subagent for investigation or fixes, with flags --background, --wait, --resume, and --fresh, plus --model and --effort overrides. /codex:transfer creates a persistent Codex thread from the current Claude Code session and prints a 'codex resume <session-id>' command. The remaining commands, /codex:status, /codex:result, and /codex:cancel, manage background jobs. Each command has a narrow job, which keeps the plugin predictable, though the number of flags may take a read of the README to internalize.
How the transfer and session mechanics work
The transfer command relies on Codex's external-agent session importer. The plugin's SessionStart hook supplies the current transcript path automatically, so /codex:transfer does not need manual input in the normal case. A --source flag overrides the path, but the source must be under ~/.claude/projects. The README notes that older Codex versions without session import support must be upgraded first. This is a concrete constraint: the transfer is not a generic file importer, it follows the same conversion rules as importing Claude history in the Codex App, and it creates visible turns that can be continued in the App or TUI. That means the transferred session is not a black box, you can reopen it later with codex resume.
Installation and setup: real commands and a readiness check
Installation is a sequence of Claude Code slash commands. First add the marketplace: /plugin marketplace add openai/codex-plugin-cc. Then install the plugin: /plugin install codex@openai-codex. Reload plugins with /reload-plugins, then run /codex:setup. That setup command checks whether Codex is installed and authenticated. If Codex is missing and npm is available, it can offer to install it. The manual fallback is npm install -g @openai/codex, and if Codex is installed but not logged in, run !codex login. The README gives a simple first run: /codex:review --background, then /codex:status, then /codex:result. Node.js 18.18 or later is a hard requirement. The setup command also manages an optional review gate with --enable-review-gate and --disable-review-gate, though the README truncates before explaining what the gate enforces.
Where the plugin is the wrong tool
The read-only commands are the safest use, but they are also the most limited. /codex:review is not steerable and takes no custom focus text, so you cannot ask it to look at a specific risk area. /codex:adversarial-review accepts focus text but still never modifies code. If your goal is to get a fix applied, you must use /codex:rescue, which does make changes and therefore carries the usual risk of an AI patch touching files you did not intend. The README warns that multi-file reviews and rescue tasks can take a long time and recommends background execution, so this is not a tool for a quick pre-commit check unless you are comfortable waiting or polling /codex:status. Also, the plugin depends on a ChatGPT subscription or OpenAI API key, and usage counts against Codex usage limits, so it is not free to run at scale.
A real alternative: running Codex CLI directly
The obvious alternative is to skip the plugin and run Codex on its own. The Codex CLI, installed via npm install -g @openai/codex, gives you the full /review command and interactive coding sessions without Claude Code in the middle. The difference is in workflow: the plugin keeps your context in Claude Code and lets you delegate or transfer a session, while the CLI requires you to move your working state manually. The plugin also adds the /codex:transfer command, which the plain CLI does not have because it is specifically about importing a Claude Code transcript. If you only want a Codex review, the CLI is simpler and has fewer moving parts. If you want to stay inside Claude Code and hand off a debugging conversation, the plugin is the bridge.
Maintenance, license, and what to verify before adopting
The repository is under Apache-2.0, which permits commercial use and modification with attribution, though this is not legal advice. The plugin is actively maintained, with releases v1.0.4, v1.0.5, and v1.0.6 between April and July 2026. The last push was July 8, 2026, so the project is not dormant. The maintenance cost for a user is mostly keeping the plugin updated via the marketplace and ensuring your Codex CLI version stays current, especially for /codex:transfer, which requires session import support. The README does not document an uninstall procedure, so you should check the marketplace plugin management commands in Claude Code before committing. Before adoption, verify that your Node.js is 18.18 or later, that /codex:setup reports Codex as ready, and that you understand the review gate behavior, which the README does not fully explain in the visible portion.
Editorial conclusion
Adopt this plugin if you are a Claude Code user who wants a second opinion from Codex on the same repo, especially for read-only reviews or background delegation. Skip it if you have no ChatGPT subscription or OpenAI API key, if you are on Node.js below 18.18, or if you need a steerable review that also fixes code; /codex:adversarial-review is read-only and /codex:rescue is the only path to changes. Before relying on it, verify that your Codex CLI is installed and logged in via /codex:setup, confirm the plugin's review gate behavior in your repo, and check that your Codex version supports session import if you plan to use /codex:transfer.
Community notes