Codex Plugin for Claude Code: How to Use OpenAI's Codex From Inside Claude Code — Complete Guide (1.5K+ Stars Today)

Why the Codex Plugin Is Trending on GitHub

If you've been following developer tools lately, you've probably noticed something incredible on GitHub Trending today. OpenAI's codex-plugin-cc repository is exploding with over 1,532 stars in a single day and 25,900+ total stars, making it one of the fastest-growing developer tools of 2026.

But what makes this plugin so special? For the first time, developers using Claude Code — Anthropic's popular AI coding assistant — can seamlessly integrate OpenAI's Codex into their existing workflow. No context switching. No copy-pasting between tools. Just one unified, powerful coding environment.

This isn't just another plugin. It's a paradigm shift in how developers think about AI-assisted coding. Instead of choosing between Claude Code and Codex, you can now use both — leveraging the strengths of each AI model for different tasks.

What Is the Codex Plugin for Claude Code?

The Codex plugin is an official integration from OpenAI that bridges Claude Code and Codex. It wraps the Codex app server and uses your local Codex CLI installation, meaning all your existing configurations, authentication, and preferences carry over automatically.

Think of it as having two expert developers sitting side by side in your terminal. Claude Code excels at understanding complex codebases and providing contextual guidance. Codex brings its own strengths in code generation, debugging, and optimization. Together, they form a powerful duo that can tackle virtually any coding challenge.

Key Features at a Glance

• Code Reviews: Get instant, AI-powered code reviews on your uncommitted changes or entire branches
• Adversarial Reviews: Challenge your design decisions and pressure-test your assumptions
• Task Delegation: Hand off bugs, refactoring tasks, or investigations to Codex
• Session Transfer: Seamlessly move conversations between Claude Code and Codex
• Background Processing: Run long-running reviews and tasks without blocking your workflow
• Review Gates: Automatically run Codex reviews before Claude Code finishes its work

How to Install the Codex Plugin

Installation takes less than a minute. Here's the complete setup process:

# Step 1: Add the OpenAI marketplace
/plugin marketplace add openai/codex-plugin-cc

# Step 2: Install the plugin
/plugin install codex@openai-codex

# Step 3: Reload plugins
/reload-plugins

# Step 4: Run setup
/codex:setup

The /codex:setup command checks whether Codex is installed and authenticated. If it's missing, the plugin can automatically install it via npm. If you prefer manual installation:

npm install -g @openai/codex

Once installed, you'll need to authenticate with either a ChatGPT subscription or an OpenAI API key:

!codex login

Core Commands and Workflows

1. Code Review with /codex:review

The most straightforward command. Run a read-only Codex review on your current work:

# Review uncommitted changes
/codex:review

# Review branch against main
/codex:review --base main

# Run in background (recommended for large changes)
/codex:review --background

This gives you the same quality of code review as running /review directly in Codex. It's perfect for catching bugs, security issues, and performance problems before they reach production.

2. Adversarial Review with /codex:adversarial-review

This is where things get interesting. Unlike regular reviews that focus on code quality, adversarial reviews challenge your architectural decisions:

# Challenge the entire approach
/codex:adversarial-review

# Focus on specific concerns
/codex:adversarial-review --base main challenge whether this was the right caching and retry design

# Look for race conditions
/codex:adversarial-review --background look for race conditions and question the chosen approach

Use this before shipping critical features. It pressure-tests your assumptions about tradeoffs, failure modes, and whether a different approach would have been safer or simpler.

3. Task Delegation with /codex:rescue

When you're stuck on a bug or need a second opinion, hand it off to Codex:

# Investigate failing tests
/codex:rescue investigate why the tests started failing

# Fix a specific issue
/codex:rescue fix the failing test with the smallest safe patch

# Use a faster, cheaper model
/codex:rescue --model gpt-5.4-mini --effort medium investigate the flaky integration test

# Quick fix with Spark model
/codex:rescue --model spark fix the issue quickly

You can also delegate tasks in natural language:

Ask Codex to redesign the database connection to be more resilient.

4. Session Transfer with /codex:transfer

Started a debugging session in Claude Code and want to continue in Codex? Transfer the entire conversation:

/codex:transfer

This creates a persistent Codex thread and prints a codex resume <session-id> command. The transfer uses Codex's external-agent session importer, preserving all context and conversation history.

Real-World Example: Debugging a Production Issue

Let's walk through a realistic scenario. You're working on a payment processing service, and the integration tests are failing intermittently in CI. Here's how you'd use the Codex plugin:

# Step 1: Get Claude Code's perspective on the issue
# (You're already in Claude Code, so ask directly)
"The integration tests for payment processing are flaky. 
They pass locally but fail 30% of the time in CI."

# Step 2: Get Codex's review of your recent changes
/codex:review --background

# Step 3: Check status
/codex:status

# Step 4: Get results
/codex:result

# Step 5: If Codex found something, great! If not, delegate investigation
/codex:rescue --background investigate why the payment integration tests 
are flaky in CI but pass locally. Focus on race conditions, 
timing issues, and external dependencies.

# Step 6: Check back later
/codex:status
/codex:result

In this workflow, you leveraged both AI assistants. Claude Code provided initial context and understanding. Codex reviewed your code and then conducted a deep investigation. You never left your terminal, never copied code between tools, and got two expert opinions on a complex problem.

Advanced Configuration

The plugin respects your existing Codex configuration. You can customize the default model and reasoning effort by creating a .codex/config.toml file in your project root:

model = "gpt-5.4-mini"
model_reasoning_effort = "high"

Configuration is loaded in this order:

• User-level config in ~/.codex/config.toml
• Project-level overrides in .codex/config.toml
• Project-level overrides only load when the project is trusted

Enabling the Review Gate

For maximum code quality, you can enable an automatic review gate:

/codex:setup --enable-review-gate

When enabled, the plugin runs a Codex review every time Claude Code is about to finish. If issues are found, the stop is blocked so Claude can address them first. This creates a powerful feedback loop, but be warned: it can consume usage limits quickly if not monitored.

Key Benefits of Using the Codex Plugin

• No Context Switching: Stay in Claude Code while accessing Codex's capabilities
• Complementary Strengths: Use Claude Code for understanding and Codex for execution
• Background Processing: Run long tasks without blocking your workflow
• Session Continuity: Transfer conversations between tools without losing context
• Automatic Configuration: Uses your existing Codex setup — no duplicate configuration
• Review Gates: Automatically catch issues before they're committed
• Cost Control: Choose different models and effort levels for different tasks
• Open Source: Fully transparent and customizable

Performance and Cost Considerations

The plugin uses your local Codex CLI, so performance depends on your machine and network connection. Multi-file reviews can take time, which is why background processing is recommended:

• Small changes (1-3 files): 10-30 seconds
• Medium changes (4-10 files): 30-90 seconds
• Large changes (10+ files): 2-5 minutes

Costs are billed to your ChatGPT subscription or OpenAI API account, the same as using Codex directly. You can control costs by:

• Using --model spark or gpt-5.4-mini for faster, cheaper tasks
• Adjusting model_reasoning_effort in your config
• Being selective about when to enable the review gate

Common Questions and Troubleshooting

Do I Need a Separate Codex Account?

No. If you're already signed into Codex on your machine, the plugin uses that authentication. If you're new to Codex, you'll need either a ChatGPT subscription (including the free tier) or an OpenAI API key. Run /codex:setup to check your status.

Does the Plugin Use a Separate Runtime?

No. The plugin delegates through your local Codex CLI and app server. This means it uses the same installation, authentication, repository checkout, and machine-local environment you'd use if running Codex directly.

Can I Keep Using My Existing API Key?

Yes. The plugin respects your existing Codex configuration, including API keys and custom base URLs. If you need to change the OpenAI endpoint, set openai_base_url in your Codex config.

What If a Task Takes Too Long?

Use /codex:cancel to stop a running task, or check status with /codex:status. For long-running work, always use the --background flag so you can continue working in Claude Code.

Can I Resume a Codex Session Later?

Yes. When you run /codex:result or /codex:status, you'll get a session ID. Use codex resume <session-id> to reopen that session directly in Codex and continue the work there.

The Future of AI-Assisted Development

The Codex plugin represents something bigger than just a tool integration. It's a glimpse into the future of software development, where multiple AI assistants work together, each contributing their unique strengths to help developers ship better code faster.

We're moving away from the "one AI to rule them all" mindset toward a more nuanced approach: use the right AI for the right task. Claude Code for understanding complex codebases and providing contextual guidance. Codex for code generation, debugging, and optimization. Together, they're more powerful than either alone.

With over 25,000 stars and growing, this plugin is clearly resonating with developers. It solves a real pain point: the friction of switching between tools, losing context, and managing multiple workflows. By bringing Codex into Claude Code, OpenAI has made it easier than ever to get the best of both worlds.

Getting Started Today

Ready to try it yourself? Here's your action plan:

1. Install the plugin with the commands shown above
2. Run /codex:setup to verify everything is working
3. Try a simple review: /codex:review --background
4. Check the results: /codex:status then /codex:result
5. Experiment with adversarial reviews and task delegation
6. Customize your configuration for your workflow

The Codex plugin is free to use (billed through your existing ChatGPT or API subscription), open source, and actively maintained by OpenAI. It's one of the most exciting developer tools to emerge in 2026, and it's available right now.

Whether you're debugging a tricky production issue, reviewing a pull request, or just want a second AI opinion on your code, the Codex plugin for Claude Code has you covered. Install it today and experience the future of AI-assisted development.