Codex Plugin For Claude Code

中文 | English

Summary

This source captures openai/codex-plugin-cc as a bridge plugin that lets Claude Code expose Codex-native review and delegation flows through its own slash-command surface, while still reusing the same local Codex install, auth state, config, and resumable sessions.

Source

• Raw file: raw/openaicodex-plugin-cc Use Codex from Claude Code to review code or delegate tasks..md
• Translated raw file: raw/openaicodex-plugin-cc Use Codex from Claude Code to review code or delegate tasks..zh.md
• Original URL: https://github.com/openai/codex-plugin-cc
• Ingest date: 2026-04-17

Key Contributions

• Exposes Codex review as Claude Code-native commands: standard review, adversarial review, delegated rescue tasks, and background-job status/result/cancel flows.
• Packages Codex delegation as a plugin install plus setup check, instead of requiring users to hand-wire app-server usage or invent their own shell wrappers.
• Reuses local Codex CLI auth, config.toml, session resume, and machine-local repository state rather than spinning up a separate runtime.
• Adds an optional stop-hook review gate, which turns Codex into a blocking verification surface inside Claude's own workflow.

Strongest Claims

• Cross-harness interoperability becomes much more usable when it is shipped as a native command surface rather than a README-only integration recipe.
• Review and delegation are not separate ecosystems here: the same Codex runtime can act as a read-only reviewer, an adversarial challenger, or a background worker from inside Claude Code.
• Review gates are powerful but operationally risky because they can create a long-running Claude/Codex loop that burns through usage quickly if left unattended.

Related Pages

• Codex Best Practices
• Codex Subagents
• Compound Engineering Plugin
• Claude Code Best Practice Repository
• Agent-First Repositories
• Codex Operating Practices