OpenAI brings Codex directly into Claude Code

openai/codex-plugin-cc is an official open-source plugin from OpenAI that lets you run Codex right inside Anthropic's Claude Code CLI. If you already live in Claude Code for daily development, this gives you a second pair of eyes without forcing you to switch windows or copy-paste context.

Under the hood, it doesn't spin up a heavy separate runtime. Instead, it delegates calls straight through your local Codex CLI and background app server. That means whatever local auth, environment variables, configs, and MCP integrations you already set up for Codex are reused automatically.

What it does

The plugin covers three primary scenarios: standard code review, adversarial review, and task delegation.

For a regular pass, running /codex:review checks your uncommitted changes or compares your branch against base with --base main. Since scanning multiple files takes time, you can append --background and inspect progress later with /codex:status and /codex:result.

The real standout feature is /codex:adversarial-review. Instead of just checking syntax and formatting, it aggressively stress-tests your architectural decisions. You can pass custom focus prompts to challenge retry policies, probe race conditions, or check rollback safety.

When Claude gets stuck on a bug, /codex:rescue hands the wheel to a Codex subagent to investigate failures or apply targeted patches. It lets you swap models on the fly—like using spark or gpt-5.4-mini—and supports resuming previous threads. You can also run /codex:transfer to export your entire Claude conversation directly into a persistent Codex session.

How to get it running

You need Node.js 18.18 or higher, along with either an OpenAI API key or a ChatGPT account (free tiers are fully supported). The usage simply draws against your regular Codex limits.

Inside Claude Code, run /plugin marketplace add openai/codex-plugin-cc to add the repo, followed by /plugin install codex@openai-codex and /reload-plugins.

Once loaded, type /codex:setup to verify your installation. If the CLI is missing, it will offer to install it, or you can run npm install -g @openai/codex manually and log in using !codex login. After that, your slash commands and the codex:codex-rescue agent are ready to roll.

Why it matters

Cross-agent verification has always been one of the best ways to catch subtle hallucinations, but context-switching across terminal windows killed the habit for most people. OpenAI meeting developers inside their existing Anthropic workflow is a genuinely pragmatic move.

If you use Claude Code daily, this is an immediate no-brainer to install for a second opinion on mission-critical PRs. Just keep an eye on your usage limits when delegating multi-file background investigations.