OpenAI Releases New Plugin, Integrating Codex into Claude Code
On March 31, 2026, OpenAI announced a new open-source plugin aimed at integrating its powerful code generation and understanding model, Codex, into the Claude Code development tool. The plugin project, named codex-plugin-cc, allows developers to leverage the power of Codex for code analysis and collaboration without leaving their current working environment.
Core Features: Three Review and Collaboration Commands
The plugin introduces three core commands to meet the needs of different development scenarios:
/codex:review (Standard Review): Performs a standard, read-only code review. This feature uses Codex to analyze code for syntax, style, and potential errors, providing routine improvement suggestions. It serves as a fundamental tool for ensuring code quality in daily development.
/codex:adversarial-review (Adversarial Review): Conducts a deeper, adversarial review. Unlike the standard review, this mode prompts Codex to actively question the implementation, logical assumptions, and potential security vulnerabilities of the code. It is particularly useful for high-risk tasks such as system migrations, changes to permission models, Infrastructure as Code (IaC) scripts, and large-scale refactoring, helping to uncover deep-seated issues in the design logic.
/codex:rescue (Task Handoff): Directly transfers control of the current task to Codex. When developers hit a technical roadblock or want AI to complete subsequent coding, they can use this command to achieve a seamless handover between human and machine collaboration.
Installation and Configuration Guide
Integrating this plugin requires two prerequisites: a valid ChatGPT account or OpenAI API key, and Node.js version 18.18 or higher installed in your local environment.
The complete installation and initialization process can be completed with the following commands:
/plugin marketplace add openai/codex-plugin-cc
/plugin install codex@openai-codex
/codex:setup
For users who have not yet configured the Codex environment locally, you first need to perform a global installation via the Node.js package manager and complete the login authentication:
Install Codex CLI globally
npm install -g @openai/codex
Log in to Codex
!codex login
Workflow and Advanced Usage
The plugin supports asynchronous processing, allowing time-consuming review tasks to run in the background. These tasks can be managed with a set of accompanying commands:
- Run in Background:
/codex:review --background
- Check Status:
/codex:status
- Get Result:
/codex:result
- Cancel Task:
/codex:cancel
The recommended usage pattern is to perform a standard review on all code changes, enable adversarial review for modifications involving core logic or high-risk modules, and use the task handoff feature to seek direct AI assistance when encountering difficulties.
Additionally, the plugin offers an optional “Review Gate” feature, which can be enabled with the following command:
/codex:setup –enable-review-gate
Once enabled, this feature prevents Claude Code from exiting before a Codex review task is complete. However, the official documentation warns that this feature could lead to prolonged cyclical calls between Claude Code and Codex, potentially consuming API quotas rapidly. Users are advised to use it with caution after careful evaluation.
How It Works
According to the source code released on GitHub, this plugin is technically a front-end adapter. It does not contain a standalone Codex runtime. Instead, it communicates with the Codex application server by invoking the locally installed Codex Command-Line Interface (CLI). This design cleverly reuses the user’s existing local Codex authentication, environment configuration, and MCP (Model and Compute Policies) settings, ensuring consistency and security. In essence, it is an encapsulation and invocation of local Codex capabilities from within Claude Code.
The plugin is open-source on GitHub, and the project repository can be found at: https://github.com/openai/codex-plugin-cc.
