Claude Proxy Server

A TypeScript proxy server that implements the Claude API and forwards requests to configurable backend providers like Qwen, OpenAI, or any OpenAI-compatible API.

Features

• Claude API Compatible: Implements the Anthropic Claude API (/v1/messages)
• Multiple Providers: Supports Qwen (DashScope) and OpenAI out of the box
• Extensible: Easy to add new providers
• Streaming Support: Full SSE streaming with proper event transformation
• Tool Use: Supports function/tool calling
• Model Mapping: Automatically maps Claude model names to provider models

Installation

npm install

Configuration

Copy the example environment file and configure your providers:

cp .env.example .env

Environment Variables

Variable	Description	Default
PORT	Server port	3000
HOST	Server host	0.0.0.0
DEFAULT_PROVIDER	Default provider to use	qwen
PROXY_API_KEY	API key for authenticating requests (optional)	-
QWEN_API_KEY	Qwen/DashScope API key	-
QWEN_BASE_URL	Qwen API base URL	https://dashscope.aliyuncs.com/compatible-mode/v1
OPENAI_API_KEY	OpenAI API key	-
OPENAI_BASE_URL	OpenAI API base URL	https://api.openai.com/v1
MODEL_MAPPINGS	JSON object mapping Claude models to provider models	See below
LOG_LEVEL	Logging level (debug, info, warn, error)	info

Model Mappings

Default model mappings:

{
  "claude-sonnet-4-5-20250514": "qwen-plus",
  "claude-sonnet-4-20250514": "qwen-plus",
  "claude-3-5-sonnet-20241022": "qwen-plus",
  "claude-3-5-haiku-20241022": "qwen-turbo",
  "claude-3-opus-20240229": "qwen-max",
  "claude-3-sonnet-20240229": "qwen-plus",
  "claude-3-haiku-20240307": "qwen-turbo"
}

Usage

Development

npm run dev

Production

npm run build
npm start

Using with Claude Code

Configure Claude Code to use this proxy:

export ANTHROPIC_BASE_URL=http://localhost:3000
export ANTHROPIC_API_KEY=your-proxy-api-key  # If PROXY_API_KEY is set

API Endpoints

POST /v1/messages

Claude Messages API compatible endpoint.

Headers:

• Content-Type: application/json
• x-api-key: <your-api-key> (if PROXY_API_KEY is configured)
• anthropic-version: 2023-06-01 (optional)

Request Body:

{
  "model": "claude-3-5-sonnet-20241022",
  "max_tokens": 1024,
  "messages": [
    {"role": "user", "content": "Hello!"}
  ],
  "stream": true
}

GET /health

Health check endpoint.

Adding New Providers

1. Create a new provider class in src/providers/:

import { BaseProvider } from './base.js';

export class MyProvider extends BaseProvider {
  readonly name = 'myprovider';

  async createCompletion(request) {
    // Implementation
  }

  async *createStreamingCompletion(request) {
    // Implementation
  }
}

export function createMyProvider(config) {
  return new MyProvider(config);
}

2. Register the provider in src/providers/index.ts:

import { createMyProvider } from './myprovider.js';
providerRegistry.register('myprovider', createMyProvider);

3. Add configuration in .env:

MYPROVIDER_API_KEY=your-key
MYPROVIDER_BASE_URL=https://api.example.com/v1

Project Structure

src/
├── index.ts              # Entry point
├── server.ts             # Express server setup
├── types/                # TypeScript type definitions
│   ├── claude.ts         # Claude API types
│   ├── openai.ts         # OpenAI-compatible types
│   └── common.ts         # Shared types
├── providers/            # Provider implementations
│   ├── base.ts           # Base provider class
│   ├── registry.ts       # Provider registry
│   ├── qwen.ts           # Qwen provider
│   └── openai.ts         # OpenAI provider
├── transformers/         # Request/response transformers
│   ├── request/          # Claude → OpenAI
│   ├── response/         # OpenAI → Claude
│   └── streaming/        # Stream transformation
├── routes/               # Route handlers
│   └── messages.ts       # /v1/messages handler
├── middleware/           # Express middleware
│   ├── auth.ts           # Authentication
│   └── error-handler.ts  # Error handling
└── utils/                # Utilities
    ├── config.ts         # Configuration
    └── logger.ts         # Logging

License

MIT