Claude Code now supports a Codex plugin that lets you call Codex without leaving your existing workflow. Once you install the plugin from https://github.com/openai/codex-plugin-cc and authenticate Codex (you need a ChatGPT subscription or an OpenAI API key plus Node.js 18.18+), you get three main commands: • /codex:review for a standard read-only review • /codex:adversarial-review for a challenge-style review that pokes at hidden assumptions • /codex:rescue to hand off a stalled task entirely to Codex Longer tasks run in the background with /codex:status, /codex:result, and /codex:cancel. To set it up, run /codex:setup once you’ve installed and authenticated Codex locally. The plugin delegates to your existing Codex CLI and server, so it respects your local auth, config files, and any MCP setup you already have. In practice, you’d run /codex:review as your go-to second pass. For anything high stakes—migrations, auth changes, infrastructure scripts, major refactors—you switch to /codex:adversarial-review to catch subtle flaws. If a thread stalls or you want Codex to take full control, /codex:rescue jumps in. There’s also an optional “review gate” you can enable to force a Codex review before Claude Code finishes—but beware: it can create tight loops that burn your usage limits fast.