OpenAI Embeds Codex as a Plugin Inside Anthropic’s Claude Code

OpenAI officially released a plugin that brings its Codex model into Anthropic’s Claude Code environment. The repository is named openai/codex-plugin-cc, where “cc” stands for Claude Code.

What the plugin can do

After installation, Claude Code gains a set of commands prefixed with /codex:: /codex:review – performs a read‑only code review, identical to the native Codex review experience. /codex:adversarial-review – a tougher review that challenges design choices, useful for high‑risk scenarios such as database migrations or permission changes. /codex:rescue – delegates a task to Codex, e.g., /codex:rescue investigate why the build is failing in CI. /codex:status and /codex:result – query the status and output of background tasks. /codex:cancel – cancels a running background task. /codex:setup – checks environment configuration and ensures the Codex CLI is ready.

Installation in three steps

/plugin marketplace add openai/codex-plugin-cc
/plugin install codex@openai-codex
/codex:setup

The prerequisites are a ChatGPT subscription (free tier works) or an OpenAI API key, and Node.js ≥ 18.18. If the Codex CLI is not yet installed, the setup command prompts to run npm install -g @openai/codex.

How it works

The plugin is essentially a thin wrapper that calls the locally installed Codex CLI and Codex App Server, reusing existing authentication, configuration files, environment variables, and MCP settings. It reads the same ~/.codex/config.toml or project‑level .codex/config.toml files, making the integration lightweight without adding a separate runtime.

Recommended usage patterns

For everyday code changes, run /codex:review to get a second pair of eyes.

For high‑risk changes, use /codex:adversarial-review to let Codex “find flaws”.

When stuck or needing a new approach, invoke /codex:rescue to let Codex investigate.

For large code reviews, add the --background flag so the task runs in the background, then check progress with /codex:status and retrieve results with /codex:result.

An experimental “Review Gate” feature can be enabled via /codex:setup --enable-review-gate. It forces Codex to review before Claude Code replies, but OpenAI warns it may cause infinite loops between the two AIs and quickly consume quota.

Why OpenAI did this

From a business perspective, Claude Code enjoys strong popularity and stickiness among developers. Rather than waiting for users to switch tools, OpenAI delivers Codex directly where users already work.

This mirrors Microsoft’s strategy of moving Office to the iPad: “Where the user is, the service should be.” Since usage of the plugin counts toward a user’s Codex quota, any invocation—whether in OpenAI’s own UI or a competitor’s—benefits OpenAI.

Technically, the move signals a trend toward composable AI programming assistants. Just as IDEs support a rich plugin ecosystem, AI coding tools are becoming modular, allowing developers to combine capabilities from different providers.

Project address: https://github.com/openai/codex-plugin-cc