Clauxy

Use Claude Code with your existing AI subscriptions.


> Note: Clauxy is in early development. GitHub Copilot with Anthropic models is stable and recommended. ChatGPT and Google Gemini support are experimental and may have issues as providers update their APIs. Feedback and contributions are welcome!

---

Why Clauxy?

Love Claude Code but also have subscriptions to GitHub Copilot, ChatGPT Plus, or Google Gemini?

Clauxy lets you use Claude Code with the AI subscriptions you already have. It acts as a transparent proxy that translates requests between Claude Code and your preferred provider.

- GitHub Copilot — Use your Copilot subscription (includes Claude, GPT, and more)
- ChatGPT Plus/Pro — Use your $20/mo ChatGPT subscription
- Google Gemini — Use Gemini models with your Google account

No API keys needed. Just authenticate once with OAuth and you're ready to go.

---

Install

``bash
npm install -g clauxy
`

Also requires Claude Code:
`bash
npm install -g @anthropic-ai/claude-code
`

---

Quick Start

Just run:

`bash
clauxy
`

First time? Select your provider and authenticate:

That's it! Clauxy starts the proxy and launches Claude Code automatically with the correct environment variables.

---

Commands

`bash


Default: Start proxy + launch Claude Code

clauxy # Just works!
clauxy -p copilot # Use specific provider
clauxy -c # Re-authenticate before starting
clauxy --port 3000 # Use specific port (reuses existing proxy if running)

Proxy-only mode (for advanced users)

clauxy run # Start proxy without launching Claude
clauxy run --port 3001 # Proxy on specific port
clauxy run -p chatgpt # Proxy with specific provider

Configuration

clauxy model # Configure model mappings
clauxy auth list # Show configured providers
`

$3

When you specify --port, Clauxy checks if a proxy is already running on that port:
- If yes, it reuses the existing proxy (no new instance started)
- If no, it starts a new proxy on that port

This lets you run multiple Claude Code sessions sharing a single proxy:

`bash


Terminal 1

clauxy --port 3000 # Starts proxy + Claude

Terminal 2

clauxy --port 3000 # Reuses proxy, launches another Claude
`

---

Model Mapping

Claude Code requests are automatically mapped to your provider's models:

| Claude Model | Copilot | ChatGPT | Gemini |
|--------------|---------|---------|--------|
| Opus (big) | Claude Opus 4.5 | GPT-5.2 Codex | Gemini 3 Pro |
| Sonnet (mid) | Claude Sonnet 4 | GPT-5.1 Codex Max | Gemini 3 Flash |
| Haiku (small) | Claude Haiku 4.5 | GPT-5.1 Codex Mini | Gemini 2.5 Flash Lite |

Customize with clauxy model or environment variables:

`bash
CLAUXY_MODEL_BIG=gpt-4o CLAUXY_MODEL_MID=gpt-4o clauxy
`

---

How It Works

`
Claude Code ──Anthropic API──▶ Clauxy ──Provider API──▶ Your Provider
│
OAuth
▼
~/.clauxy/
`

1. Clauxy intercepts Claude Code's Anthropic API requests
2. Translates them to your provider's format (OpenAI, Gemini, etc.)
3. Returns responses in Anthropic format

Credentials are stored locally in ~/.clauxy/ with restricted permissions.

---

Troubleshooting

"Token refresh failed" or "Not authenticated"
`bash
clauxy -c
`

Port already in use
`bash
clauxy --port 3001
`

Reset everything
`bash
rm -rf ~/.clauxy
`

---

Security

- OAuth tokens stored locally with chmod 600`
- Tokens auto-refresh, no manual handling
- No data sent anywhere except your chosen provider
- Fully open source — audit the code

---

Contributing

See CONTRIBUTING.md for development setup and guidelines.

---

License

MIT