Claude Bridge

A smart, fast, and interactive proxy server that enables the Claude Code CLI to work with any OpenAI-compatible API provider.

Key Features

• 🚀 Simple & Fast: Minimal dependencies, near-instant startup.
• 🤖 Interactive Model Selection: If you don't specify a model, the bridge will automatically fetch the available models from your provider and present you with an interactive list to choose from.
• 🧠 Smart Token Auto-Detection: Automatically detects the max_tokens limit for your chosen model and sets a safe cap, preventing common errors.
• 🛠️ Full Tool Support: Seamlessly converts function-calling and tool-use formats between Claude and OpenAI.
• 🌐 Universal Compatibility: Works with OpenAI, Azure OpenAI, Ollama, OpenRouter, and any other OpenAI-compatible API.
• 🛡️ Robust & Type-Safe: Built with TypeScript for reliability and includes graceful error handling.
• 🐳 Production Ready: Full Docker support, health checks, and graceful shutdown.

Quick Start

The easiest way to get started is with npx.

1. Run the Bridge: Provide your API provider's URL and your API key. You can omit the model name to get an interactive selector.

   # Run without a model to get an interactive prompt
   npx claude-bridge -u https://api.your-provider.com/v1 -k sk-your-key
   
   # Or, specify a model directly
   npx claude-bridge -u https://api.your-provider.com/v1 -k sk-your-key -m gpt-4o
   
   # Full example with OpenRouter
   npx claude-bridge -u https://openrouter.ai/api/v1/ -k <your_api_key> -m deepseek/deepseek-chat-v3-0324:free
   
   # Example with PPInfra
   npm run dev -- -u https://api.ppinfra.com/v3/openai/ -k <your_api_key> -m deepseek/deepseek-v3-0324

2. Configure the Claude CLI: In another terminal, point the Claude CLI to the bridge:

   export ANTHROPIC_BASE_URL=http://localhost:8000
   export ANTHROPIC_AUTH_TOKEN="dummy"
   claude

That's it! The Claude CLI will now work through the bridge.

CLI Options

Flag	Alias	Description	Default
--url <string>	-u	(Required) Target API base URL.
--key <string>	-k	(Required) Your API key.
--model <string>	-m	Model name. If omitted, an interactive selector will appear.
--port <number>	-p	The port for the bridge server to run on.	8000
--host <string>		The host address for the bridge server.	localhost
--max-tokens <number>		Manually override the maximum tokens for a model.	(auto-detected)
--timeout <number>		Request timeout in seconds.	30
--log-level <level>		Set the log level (debug, info, warn, error).	info
--help	-h	Show the help message.
--version	-v	Show the version number.

How It Works

The bridge acts as an intelligent middleman:

1. Receives a request from the Claude Code CLI.
2. Converts the Claude API format (including tool-use) to the standard OpenAI format.
3. Forwards the request to your chosen API provider.
4. Converts the OpenAI response back to the Claude format.
5. Returns the final response to the Claude Code CLI.

Claude Code CLI  →  Claude Bridge  →  Any OpenAI-compatible API
 (Claude Format)   (Conversion)      (OpenAI Format)

Development

# Clone and install dependencies
git clone https://github.com/loulin/claude-bridge.git
cd claude-bridge
npm install

# Run the development server
# This will first build the project, then run it with nodemon.
npm run dev

# Manually build the project
npm run build

# Run tests
npm test

License

MIT