OpenAI Ships Official Codex Plugin for Anthropic’s Claude Code
OpenAI has released an official plugin that embeds its Codex coding agent directly inside Anthropic’s Claude Code CLI. Announced on March 30, 2026, the open-source codex-plugin-cc plugin lets developers invoke Codex for code reviews, adversarial analysis, and task delegation — all without leaving the Claude Code terminal. The move is notable because OpenAI is deliberately shipping its AI agent into a competitor’s ecosystem rather than trying to lure users away from it.
Intermediate
What the Plugin Does
The plugin, built by OpenAI engineers Dominik Kundel, Kyle Kelley, and Omid Rajabi, adds six slash commands to Claude Code:
/codex:review— Runs a standard read-only code review on uncommitted changes or branch diffs, providing the same quality as running/reviewdirectly in Codex./codex:adversarial-review— A more aggressive “devil’s advocate” review that challenges design decisions, questions trade-offs, and tests implementation assumptions. Particularly useful for migrations and authentication changes./codex:rescue— Delegates tasks directly to Codex as a subagent. Developers can hand off debugging, test fixes, or entire implementation chunks by describing the problem in natural language./codex:status,/codex:result,/codex:cancel— Background job management commands for monitoring, retrieving output, and terminating running tasks./codex:setup— Checks installation status and manages an optional review gate.
All review and rescue commands support background execution, letting developers continue working in Claude Code while Codex processes in parallel. Work delegated to Codex can also be resumed directly via codex resume <session-id> for continued development outside the plugin.
Setup and Requirements
Installation takes three commands inside Claude Code:
/plugin marketplace add openai/codex-plugin-cc
/plugin install codex@openai-codex
/reload-pluginsAfter installation, running /codex:setup verifies everything is connected. The plugin requires Node.js 18.18 or later and either a ChatGPT subscription (including the free tier) or an OpenAI API key. It wraps the locally installed Codex CLI binary and reuses existing authentication from ~/.codex/config.toml, so there is no separate account or runtime to manage. Developers can customize the default model (e.g., gpt-5.4-mini) and reasoning effort through the same configuration file.
Why OpenAI Built This
The strategic context makes this release particularly interesting. Claude Code has become the dominant agentic coding tool, reaching an estimated $2.5 billion in annualized revenue by early 2026 and accounting for roughly 135,000 daily GitHub commits — about 4% of all public commits. Codex, while growing to over 2 million weekly active users, trails in developer loyalty.
Rather than compete head-on for users, OpenAI is meeting developers where they already are. Engineer Dominik Kundel framed the approach as ecosystem openness: shipping Codex “whether that’s in our apps, in Xcode, JetBrains, OpenCode, Pi, or even Claude Code.” But the economics reveal deeper positioning — each /codex:review or /codex:rescue call consumes OpenAI API tokens, creating revenue from Claude Code’s user base without requiring full user acquisition.
Early developer reactions have been mixed but pragmatic. Game framework creator Mario Zechner called it “hilarious,” while developer Austin Wallace highlighted genuine value: running both models against the same code catches different blind spots, with Claude tending to find architectural issues and Codex catching correctness problems.
Broader Codex Updates
The Claude Code plugin arrives alongside a broader expansion of Codex’s capabilities in March 2026. OpenAI launched an enterprise plugin system allowing organizations to package workflows, app integrations, and MCP server configurations into installable bundles. Over 20 plugins are now available across the Codex app, CLI, and VS Code extension, with integrations for Figma, Notion, Gmail, Google Drive, and Slack. OpenAI also introduced Codex Security, an application-security agent that identified nearly 800 critical vulnerabilities across projects including Chromium and OpenSSL during testing on 1.2 million commits.
Related Coverage
- Unlocking Efficiency: Best Practices for Agentic Coding with Claude Code — deep dive into Claude Code workflows
- OpenAI Codex Update: Now Available to ChatGPT Plus Users — previous Codex expansion to Plus tier
- Anthropic Releases Claude Opus 4.6 with 1M Token Context Window — the model powering Claude Code