OpenAI Ships an Official Plugin for Claude Code
This one takes a second to sink in. OpenAI has an official repo called codex-plugin-cc, and what it does is let you run Codex from inside Claude Code. Type /codex:review and OpenAI's coding agent reviews the code Anthropic's agent just wrote. There's even /codex:adversarial-review, which explicitly tells Codex to attack the implementation decisions. 22.6k stars, Apache-2.0, v1.0.5 shipped June 23, and it's pulling around 450 stars a day on GitHub Trending right now.
The feature set goes beyond review. /codex:rescue hands a stuck task over to Codex in the background, /codex:transfer moves a whole session across, and there's job management to track what Codex is doing while Claude keeps working. It uses your existing local Codex CLI and auth, so a ChatGPT subscription or an API key is all you need.
Why does this matter more than another editor extension? Because it's OpenAI conceding where developers actually live. Claude Code is the shell, the harness, the place the workday happens, and rather than fight that, OpenAI shipped into it. The harness has gravity now. Models are becoming subagents you summon into someone else's terminal.
The practical upside is real too: two frontier labs' models arguing over your diff is genuinely useful. One writes, the other tries to tear it apart, and you read the argument. That's cheaper than a human reviewer and far less agreeable than asking the same model to check its own work.
https://github.com/openai/codex-plugin-cc
← Back to all articles
The feature set goes beyond review. /codex:rescue hands a stuck task over to Codex in the background, /codex:transfer moves a whole session across, and there's job management to track what Codex is doing while Claude keeps working. It uses your existing local Codex CLI and auth, so a ChatGPT subscription or an API key is all you need.
Why does this matter more than another editor extension? Because it's OpenAI conceding where developers actually live. Claude Code is the shell, the harness, the place the workday happens, and rather than fight that, OpenAI shipped into it. The harness has gravity now. Models are becoming subagents you summon into someone else's terminal.
The practical upside is real too: two frontier labs' models arguing over your diff is genuinely useful. One writes, the other tries to tear it apart, and you read the argument. That's cheaper than a human reviewer and far less agreeable than asking the same model to check its own work.
https://github.com/openai/codex-plugin-cc
Comments