You Can Now Use OpenAI’s Codex in Claude Code

OpenAI has released a Codex plugin for Claude Code, enabling developers to access Codex capabilities within Anthropic’s coding environment. The plugin integrates with the local Codex command-line interface, allowing users to run code reviews, perform adversarial analysis, and delegate tasks without leaving their existing workflow.

The tool is available via GitHub and can be installed through Claude Code’s plugin marketplace. It requires either a ChatGPT account, including the free tier, or an OpenAI API key, along with Node.js version 18.18 or later. 

The plugin uses the locally installed Codex binary and shares the same authentication, configuration, and runtime environment.

The integration introduces several commands. 

The “/codex:review” command performs a read-only review of code changes, including comparisons against a base branch. 

The “/codex:adversarial-review” command allows developers to request a more critical evaluation that challenges implementation decisions and identifies risks such as race conditions or design trade-offs. 

A third command, “/codex:rescue”, enables users to hand off tasks such as debugging or implementing fixes directly to Codex.

According to the project documentation, “This plugin is for Claude Code users who want an easy way to start using Codex from the workflow they already have.” 

The plugin also supports background execution for longer tasks, with commands to check status, retrieve results, or cancel jobs.

The system works by delegating requests through the Codex CLI and app server running locally on the developer’s machine. “It is not a separate runtime. It is Codex, just invoked from inside Claude Code,” the document stated.

This is a real dario and sam hand holding moment https://t.co/hfrMJ7YRIX pic.twitter.com/THo6THub72

— Manthan Gupta (@manthanguptaa) March 30, 2026

This design allows the plugin to reuse existing configurations, including model selection and reasoning settings defined in Codex configuration files.

An optional review gate feature can be enabled to enforce a Codex review before Claude Code completes a task. However, the documentation warns that this may create extended feedback loops and increase usage.