Recently, OpenAI officially stepped in to develop a plugin called openai/codex-plugin-cc for Claude Code. It's specifically designed for Claude Code users, and its main purpose is to integrate OpenAI Codex CLI directly into your Claude Code working environment.

Imagine you're in the middle of a coding conversation in Claude Code, and suddenly you need to use Codex for in-depth code review, or delegate a complex debugging task. Without this plugin, you would typically need to stop the current session, switch to the Codex environment, re-establish the context, and then execute the task. With this plugin, you can seamlessly accomplish all of this in the same environment, greatly improving development efficiency.

How to Install the Codex Plugin

The installation process is very simple. Just add the marketplace source in Claude Code and install directly. Please execute the following three commands in order:

/plugin marketplace add openai/codex-plugin-cc
/plugin install codex@openai-codex
/reload-plugins

Detailed Explanation of Codex Plugin Commands

This plugin provides 8 main commands, covering four functional categories: code review, task delegation, task management, and configuration.

Code Review Commands

/codex:review

Standard code review, performs a read-only review of current code changes, providing the same quality of review results as running /review directly in Codex.

Parameter Options:

  • --base <ref>: Specify a base branch for branch review
  • --scope <auto|working-tree|branch>: Specify the review scope
  • --wait: Wait in foreground for completion
  • --background: Run in background

Usage Examples:

/codex:review # Review current uncommitted changes
/codex:review --base main # Review changes relative to main branch
/codex:review --background # Run review in background

This is a read-only operation suitable for quality checks before committing, and it will not modify code.

/codex:adversarial-review

Adversarial review, runs a guided review process that proactively questions code implementation and design choices, used for stress-testing assumptions, trade-offs, and potential failure modes.

Parameter Options:

  • --base <ref>: Specify base branch
  • --scope <auto|working-tree|branch>: Specify review scope
  • --wait: Wait in foreground for completion
  • --background: Run in background
  • [focus text]: Additional focus text

Usage Examples:

/codex:adversarial-review # Adversarial review of current changes
/codex:adversarial-review --base main challenge whether this was the right caching design
/codex:adversarial-review --background look for race conditions

This is a read-only review mechanism focused on uncovering design choices, trade-offs, and hidden assumptions. While it won't directly fix code, it supports targeted investigation through custom focus text, making it perfect for in-depth design review before release.

The previous two commands are only for reviewing and won't modify code. If you want AI to help fix issues together, use the commands below.

Task Delegation Commands

/codex:rescue

Task delegation and programming, delegates tasks to Codex through the codex:codex-rescue sub-agent, supporting bug investigation, fix attempts, and continuing previous tasks.

Parameter Options:

  • --background: Run in background
  • --wait: Wait in foreground for completion
  • --resume: Continue previous Codex thread
  • --fresh: Start new thread
  • --model <model|spark>: Specify model
  • --effort <none|minimal|low|medium|high|xhigh>: Reasoning effort level
  • [task description]: Task description

Usage Examples:

/codex:rescue investigate why the tests started failing
/codex:rescue fix the failing test with the smallest safe patch
/codex:rescue --resume apply the top fix from the last run
/codex:rescue --model gpt-5.4-mini --effort medium investigate the flaky test
/codex:rescue --model spark fix the issue quickly
/codex:rescue --background investigate the regression

This command is perfect for investigating complex bugs, attempting code fixes, continuing unfinished Codex tasks, or using smaller, faster models for rapid iteration.

/codex:transfer

Session transfer, creates a persistent Codex thread from the current Claude Code session and outputs the codex resume <session-id> command.

Parameter Options:

  • --source <claude-jsonl>: Manually specify Claude session file path

Usage Examples:

/codex:transfer
/codex:transfer --source ~/.claude/projects/-Users-me-repo/<session-id>.jsonl

This command is suitable for scenarios where you want to seamlessly transfer the current context to Codex for continued processing after starting a debugging or implementation conversation in Claude Code.

Task Management Commands

/codex:status

View task status, displays currently running and recent Codex tasks in the current repository.

Parameter Options:

  • [job-id]: View detailed information for a specific task
  • --wait: Wait for task to reach terminal state
  • --timeout-ms <ms>: Wait timeout in milliseconds
  • --all: Display all tasks (not limited to current session)

Usage Examples:

/codex:status # Display task list for current session
/codex:status task-abc123 # Display detailed information for specific task
/codex:status --wait task-abc123 # Wait for task completion

The output includes task ID, type, status and stage, runtime or duration, Codex session ID, task summary, and follow-up action commands.

/codex:result

View task results, displays the final stored output of completed tasks, and includes the Codex session ID for reopening the task directly in Codex.

Parameter Options:

  • [job-id]: Specify task ID

Usage Examples:

/codex:result # Display latest task result
/codex:result task-abc123 # Display specific task result

The output includes task ID and status, complete result payload (verdict, summary, findings, details, artifacts, next_steps), file paths and line numbers, error messages (if any), and follow-up action commands.

/codex:cancel

Cancel task, used to terminate active background Codex tasks. If no parameter is passed, it cancels the latest task by default.

Parameter Options:

  • [job-id]: Specify task ID to cancel

Usage Examples:

/codex:cancel # Cancel latest task
/codex:cancel task-abc123 # Cancel specific task

Configuration Commands

/codex:setup

Setup and configuration, used to check if Codex is installed and authenticated. If Codex is missing and npm is available locally, this command can also provide installation guidance.

Parameter Options:

  • --enable-review-gate: Enable review gate
  • --disable-review-gate: Disable review gate
  • --json: Output in JSON format

Usage Examples:

/codex:setup # Check Codex status
/codex:setup --enable-review-gate # Enable review gate
/codex:setup --disable-review-gate # Disable review gate

When the review gate feature is enabled, the plugin uses Stop hooks to run targeted Codex reviews based on Claude's responses. If issues are found during review, it will prevent the session from stopping so that Claude can prioritize resolving these issues first. Note that this may trigger long-running Claude/Codex cycles, quickly consuming tokens, so avoid enabling it unless necessary.

Conclusion

If you're already using Claude Code, I strongly recommend installing this OpenAI official plugin right away. It can take your code quality to the next level.

ChatGPT