OpenAI-to-Claude API Proxy Service

High-performance proxy service that converts OpenAI API to Anthropic API compatible format. Allows developers to seamlessly call OpenAI models using existing Anthropic client code.

中文版本

🌟 Core Features

• ✅ Seamless Compatibility: Call OpenAI models using standard Anthropic clients
• ✅ Full Functionality: Supports text, tool calls, streaming responses, and more
• ✅ Intelligent Routing: Automatically selects the most suitable OpenAI model based on request content
• ✅ Hot Reload: Automatically reloads configuration file changes without restarting the service
• ✅ Structured Logging: Detailed request/response logs for debugging and monitoring
• ✅ Error Mapping: Comprehensive error handling and mapping mechanisms

🚀 Quick Start

Requirements

• Python 3.11+
• uv (recommended package manager)

Install Dependencies

# Install dependencies using uv (recommended)
uv sync

Configuration

1. Copy the example configuration file:

cp config/example.json config/settings.json

2. Edit config/settings.json:

{
  "openai": {
    "api_key": "your-openai-api-key-here",  // Replace with your OpenAI API key
    "base_url": "https://api.openai.com/v1"  // OpenAI API address
  },
  "api_key": "your-proxy-api-key-here",  // API key for the proxy service
  // Other configurations...
}

Start the Service

# Development mode
uv run main.py --config config/settings.json

# Production mode
uv run main.py

Start with Docker

# Build and start the service
docker-compose up --build

# Run in background
docker-compose up --build -d

# Stop the service
docker-compose down

The service will start at http://localhost:8000.

🛠️ Usage

Claude Code Usage

This project can be used with Claude Code for development and testing. To configure Claude Code to work with this proxy service, create a .claude/settings.json file with the following configuration:

{
    "env": {
        "ANTHROPIC_API_KEY": "your-api-key",
        "ANTHROPIC_BASE_URL": "http://127.0.0.1:8000",
        "DISABLE_TELEMETRY": "1",
        "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1"
    },
    "apiKeyHelper": "echo 'your-api-key'",
    "permissions": {
        "allow": [],
        "deny": []
    }
}

Configuration Notes:

• Replace ANTHROPIC_API_KEY with your API key, in config/settings.json
• Replace ANTHROPIC_BASE_URL with the actual URL where this proxy service is running
• The apiKeyHelper with your API key, in config/settings.json

Using Anthropic Python Client

from anthropic import Anthropic

# Initialize client pointing to the proxy service
client = Anthropic(
    base_url="http://localhost:8000/v1",
    api_key="your-proxy-api-key-here"  # Use the api_key from the configuration file
)

# Send a message request
response = client.messages.create(
    model="gpt-4o",
    messages=[
        {"role": "user", "content": "Hello, GPT!"}
    ],
    max_tokens=1024
)

print(response.content[0].text)

Streaming Response

# Streaming response
stream = client.messages.create(
    model="gpt-4o",
    messages=[
        {"role": "user", "content": "Tell me a story about AI"}
    ],
    max_tokens=1024,
    stream=True
)

for chunk in stream:
    if chunk.type == "content_block_delta":
        print(chunk.delta.text, end="", flush=True)

Tool Calls

# Tool calls
tools = [
    {
        "name": "get_current_weather",
        "description": "Get the current weather for a specified city",
        "input_schema": {
            "type": "object",
            "properties": {
                "city": {
                    "type": "string",
                    "description": "City name"
                }
            },
            "required": ["city"]
        }
    }
]

response = client.messages.create(
    model="gpt-4o",
    messages=[
        {"role": "user", "content": "What's the weather like in Beijing now?"}
    ],
    tools=tools,
    tool_choice={"type": "auto"}
)

📁 Project Structure

openai-to-claude/
├── src/
│   ├── api/             # API endpoints and middleware
│   ├── config/          # Configuration management
│   ├── core/            # Core business logic
│   │   ├── clients/     # HTTP clients
│   │   └── converters/  # Data format converters
│   ├── models/          # Pydantic data models
│   └── common/          # Common utilities (logging, token counting, etc.)
├── config/              # Configuration files
├── tests/               # Test suite
├── CLAUDE.md           # Claude Code project guide
└── pyproject.toml      # Project dependencies and configuration

🔧 Configuration

Environment Variables

• CONFIG_PATH: Configuration file path (default: config/settings.json)
• LOG_LEVEL: Log level (default: INFO)

Configuration File (config/settings.json)

{
  "openai": {
    "api_key": "your-openai-api-key-here",
    "base_url": "https://api.openai.com/v1"
  },
  "server": {
    "host": "0.0.0.0",
    "port": 8000
  },
  "api_key": "your-proxy-api-key-here",
  "logging": {
    "level": "INFO"
  },
  "models": {
    "default": "Qwen/Qwen3-Coder",
    "small": "deepseek-ai/DeepSeek-V3-0324",
    "think": "deepseek-ai/DeepSeek-R1-0528",
    "long_context": "gemini-2.5-pro",
    "web_search": "gemini-2.5-flash"
  },
  "parameter_overrides": {
    "max_tokens": null,
    "temperature": null,
    "top_p": null,
    "top_k": null
  }
}

Configuration Items

• openai: OpenAI API configuration

  • api_key: OpenAI API key for accessing OpenAI services
  • base_url: OpenAI API base URL, default is https://api.openai.com/v1

• server: Server configuration

  • host: Service listening host address, default is 0.0.0.0 (listen on all network interfaces)
  • port: Service listening port, default is 8000

• api_key: API key for the proxy service, used to verify requests to the /v1/messages endpoint

• logging: Logging configuration

  • level: Log level, options are DEBUG, INFO, WARNING, ERROR, CRITICAL, default is INFO

• models: Model configuration, defines model selection for different usage scenarios

  • default: Default general model for general requests
  • small: Lightweight model for simple tasks
  • think: Deep thinking model for complex reasoning tasks
  • long_context: Long context processing model for handling long text
  • web_search: Web search model for web search, currently supports geimini

• parameter_overrides: Parameter override configuration, allows administrators to set model parameter override values in the configuration file

  • max_tokens: Maximum token count override, when set, will override the max_tokens parameter in client requests
  • temperature: Temperature parameter override, controls the randomness of output, range 0.0-2.0
  • top_p: top_p sampling parameter override, controls the probability threshold of candidate words, range 0.0-1.0
  • top_k: top_k sampling parameter override, controls the number of candidate words, range >=0

🧪 Testing

# Run all tests
pytest

# Run unit tests
pytest tests/unit

# Run integration tests
pytest tests/integration

# Generate coverage report
pytest --cov=src --cov-report=html

📊 API Endpoints

• POST /v1/messages - Anthropic Messages API
• GET /health - Health check endpoint
• GET / - Welcome page

🛡️ Security

• API key authentication
• Request rate limiting (planned)
• Input validation and sanitization
• Structured logging

📈 Performance Monitoring

• Request/response time monitoring
• Memory usage tracking
• Error rate statistics

🤝 Contributing

Issues and Pull Requests are welcome!

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgements

• claude-code-router - Very good project, many places in this project have referenced this project
• FastAPI - Modern high-performance web framework
• Anthropic - Claude AI models
• OpenAI - OpenAI API specification

🤖 Claude Code Usage

This project can be used with Claude Code for development and testing. To configure Claude Code to work with this proxy service, create a .claude/settings.json file with the following configuration:

Example Configuration

{
    "env": {
        "ANTHROPIC_API_KEY": "sk-chen0v0...",
        "ANTHROPIC_BASE_URL": "http://127.0.0.1:8100",
        "DISABLE_TELEMETRY": "1",
        "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1"
    },
    "apiKeyHelper": "echo 'sk-chen0v0...'",
    "permissions": {
        "allow": [],
        "deny": []
    }
}

Configuration Notes

• Replace ANTHROPIC_API_KEY with your actual Anthropic API key
• Replace ANTHROPIC_BASE_URL with the actual URL where this proxy service is running
• The apiKeyHelper field should also be updated with your actual API key