Claude Code Proxy

A simple HTTP proxy that wraps the Claude Code CLI so you can call it via REST APIs. Supports OpenAI and Anthropic API formats, including streaming.

Why?

Sometimes you want to hit Claude Code from tools that speak HTTP instead of running CLI commands directly. This lets you do that.

Requirements

• Node.js 20.6+
• Claude Code CLI installed and logged in

Setup

npm install

# Create .env with your API key
echo "PROXY_API_KEY=pick-something-secret" > .env

# Run it
npm run dev

The server starts on port 6789 by default.

Configuration

Set these in .env if you want to change defaults:

Variable	Default	What it does
PROXY_API_KEY	required	Your auth key for the proxy
PORT	6789	Server port
LOG_LEVEL	info	debug, info, warn, error
REQUEST_TIMEOUT_MS	300000	5 min timeout (Claude can be slow)
WORKER_CONCURRENCY	2	How many Claude processes run at once
MAX_QUEUE_SIZE	100	Requests queued before rejecting

API

All endpoints need Authorization: Bearer <your-proxy-api-key> except /health.

Health Check

GET /health

Returns status, uptime, and queue stats. No auth needed.

OpenAI Format

Works with tools that expect OpenAI's API.

# List models
curl http://localhost:6789/v1/models \
  -H "Authorization: Bearer $API_KEY"

# Chat completion
curl http://localhost:6789/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sonnet",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

# Streaming (use -N to disable buffering)
curl -N http://localhost:6789/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sonnet",
    "stream": true,
    "messages": [{"role": "user", "content": "Write a haiku"}]
  }'

Anthropic Format

curl http://localhost:6789/v1/messages \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-20250514",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Direct API

If you want more control (tool restrictions, working directory, sessions):

curl http://localhost:6789/api/run \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "List files here",
    "allowedTools": ["Bash", "Read"],
    "workingDirectory": "/some/path"
  }'

Sessions

List and delete conversation sessions:

# List sessions
curl http://localhost:6789/api/sessions \
  -H "Authorization: Bearer $API_KEY"

# Delete a session
curl -X DELETE http://localhost:6789/api/sessions/session-id \
  -H "Authorization: Bearer $API_KEY"

Testing with Insomnia

Import insomnia-collection.json for a ready-to-go collection with all the endpoints.

Note: Insomnia's default 30s timeout is too short for Claude. Go to Preferences → General → Request timeout and bump it up or set to 0 (unlimited).

Development

npm run dev      # Dev mode with hot reload
npm test         # Run tests
npm run lint     # Lint
npm run build    # Build for production

How it Works

Your App → HTTP Request → Proxy → spawns claude CLI → Response back

The proxy manages a pool of workers so multiple requests can run concurrently. Failed requests retry automatically with backoff (timeouts, rate limits). Streaming pipes chunks back as SSE.

Limitations

• Token counts are always 0 (CLI doesn't expose them)
• Some params like temperature are ignored (logged but not used)
• Sessions are in-memory only, lost on restart

License

MIT