OpenAI integrates with Claude Code
- OpenAI published an official GitHub plugin that lets Anthropic’s Claude Code invoke Codex for code reviews and delegated background jobs from the same terminal workflow. - The plugin adds slash commands including `/codex:review`, `/codex:adversarial-review`, `/codex:rescue`, `/codex:status`, and `/codex:result`, and works with a ChatGPT account or OpenAI API key. - The release lands as both Codex and Claude Code expand Model Context Protocol tooling, making cross-agent workflows easier. (openai.com) (code.claude.com)
OpenAI has released an official plugin that lets developers run Codex from inside Anthropic’s Claude Code, instead of switching between separate coding agents. (github.com) (community.openai.com) The repository, `openai/codex-plugin-cc`, says Claude Code users can use Codex for code review or hand off work as background jobs. OpenAI announced the plugin in its developer community on March 30, 2026. (github.com) (community.openai.com) The plugin ships with slash commands for a standard review, an “adversarial” review, and job control commands including status, result, and cancel. OpenAI says it works with a ChatGPT subscription, including Free, or with an OpenAI API key. (github.com) (community.openai.com) Claude Code is Anthropic’s terminal coding tool, and it can connect to outside services through the Model Context Protocol, or MCP. Anthropic describes MCP as an open standard that lets Claude Code read from and act on tools, databases, and application programming interfaces directly. (code.claude.com) OpenAI is building around the same plumbing on its side. Its Codex docs say plugins can package skills, app integrations, and MCP server configuration, and its MCP docs explain how Codex connects to external servers through shared configuration files. (openai.com 1) (openai.com 2) That is why this integration is more than a simple shortcut menu. OpenAI says the Claude Code plugin delegates through the local Codex command-line interface and Codex app server, so it uses the same authentication, environment, and MCP setup a developer already has in Codex. (community.openai.com) (openai.com) A separate Composio guide shows another route into the same idea: connecting OpenAI to Claude Code through a universal MCP server. Composio says its Rube server exposes more than 850 software-as-a-service apps and keeps large tool responses out of the model’s context window through a remote workbench. (composio.dev) That Composio setup is not an OpenAI product, but it points in the same direction as the official plugin. Instead of one assistant living in one chat window, developers are starting to chain agents, tools, and app connections inside a single coding session. (composio.dev) (openai.com) OpenAI’s own Codex pages now describe the product as a coding agent that works across the terminal, desktop app, and integrated development environments. Anthropic’s docs describe Claude Code as a tool that can pull from issue trackers, databases, design tools, and messaging systems through MCP servers. (openai.com) (code.claude.com) The immediate result is simple: a Claude Code user can ask Codex for a second opinion without leaving the workflow they already use. The bigger shift is that the leading coding agents are starting to meet inside the same toolchain. (github.com) (community.openai.com)