Claude Max API Proxy

English | 繁體中文

Use your Claude Max subscription with any OpenAI-compatible client.

Why This Exists

Claude Max ($200/month) offers unlimited access to Claude, but Anthropic restricts it to the web UI and Claude Code CLI — you can't use your subscription to power third-party tools.

This proxy works around that limitation. It spawns the real Claude Code CLI as a subprocess and exposes an OpenAI-compatible HTTP API locally. Any client that speaks the OpenAI chat completions protocol can use your Max subscription as the backend — including OpenClaw for Telegram/Discord bots.

How It Works

┌─────────────┐     HTTP      ┌──────────────────┐    spawn()    ┌───────────────┐
│  Any OpenAI  │ ──────────▶ │  Claude Max API   │ ──────────▶ │  Claude Code   │
│  compatible  │ ◀────────── │  Proxy (Express)  │ ◀────────── │  CLI (--print) │
│  client      │   SSE/JSON   │  localhost:3456   │  stream-json │               │
└─────────────┘               └──────────────────┘              └───────────────┘

No third-party servers. Everything runs locally. Requests go through Anthropic's own CLI binary — identical to you typing in your terminal.

Key Features

• OpenAI-compatible API — Drop-in replacement for any client that supports POST /v1/chat/completions
• Streaming & non-streaming — Full SSE streaming support with direct delta forwarding
• Session persistence — Conversations maintain context across messages via CLI session resume
• No turn limits — The CLI runs as many tool-call rounds as needed for complex tasks
• Activity timeout — 10-minute inactivity watchdog catches stuck processes while letting long tasks complete
• Telegram progress — Real-time progress updates showing which tools are running (optional)
• No native dependencies — Pure JS, uses child_process.spawn() with piped stdio and --output-format stream-json

Quick Start

Prerequisites

1. Claude Max subscription ($200/month) — Subscribe here
2. Claude Code CLI installed and authenticated:

   npm install -g @anthropic-ai/claude-code
   claude auth login

Install & Run

npm install -g claude-max-api-proxy
claude-max-api   # starts on http://localhost:3456

Test

# Health check
curl http://localhost:3456/health

# Chat (streaming)
curl -N -X POST http://localhost:3456/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

Building from Source

The project ships with full TypeScript source in src/.

git clone https://github.com/GodYeh/claude-max-api-proxy.git
cd claude-max-api-proxy
npm install
npm run build    # compiles src/ → dist/
npm run start    # starts the server

Configuration

Environment Variables

Variable	Required	Description
TELEGRAM_NOTIFY_ID	No	Telegram user ID for timeout notifications
DEBUG	No	Set to any value to enable request logging

Available Models

Model ID	Description
claude-opus-4	Claude Opus (most capable)
claude-sonnet-4	Claude Sonnet (balanced)
claude-haiku-4	Claude Haiku (fastest)

Full model family support with version pinning (e.g. claude-opus-4-5-20251101, claude-sonnet-4-20250514).

Auto-Start on macOS

Create ~/Library/LaunchAgents/com.claude-max-api.plist:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"
  "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
  <dict>
    <key>Label</key>
    <string>com.claude-max-api</string>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <dict>
      <key>SuccessfulExit</key>
      <false/>
    </dict>
    <key>ProgramArguments</key>
    <array>
      <string>/opt/homebrew/bin/node</string>
      <string>/opt/homebrew/lib/node_modules/claude-max-api-proxy/dist/server/standalone.js</string>
    </array>
    <key>EnvironmentVariables</key>
    <dict>
      <key>HOME</key>
      <string>/Users/YOUR_USERNAME</string>
      <key>PATH</key>
      <string>/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin</string>
    </dict>
    <key>StandardOutPath</key>
    <string>/tmp/claude-max-api.log</string>
    <key>StandardErrorPath</key>
    <string>/tmp/claude-max-api.err.log</string>
  </dict>
</plist>

Then load it:

launchctl load ~/Library/LaunchAgents/com.claude-max-api.plist

OpenClaw Integration

Add as a model provider in your openclaw.json:

{
  "models": {
    "providers": {
      "maxproxy": {
        "baseUrl": "http://127.0.0.1:3456/v1",
        "apiKey": "not-needed",
        "api": "openai-completions"
      }
    }
  }
}

When used with OpenClaw, this proxy supports all native agent features: web search, browser automation, voice messages, scheduled tasks, media attachments, and more.

API Endpoints

Endpoint	Method	Description
/health	GET	Health check
/v1/models	GET	List available models
/v1/chat/completions	POST	Chat completions (streaming & non-streaming)

Project Structure

src/
├── adapter/
│   ├── openai-to-cli.ts    # OpenAI request → CLI prompt + system prompt
│   └── cli-to-openai.ts    # CLI JSON stream → OpenAI response format
├── subprocess/
│   └── manager.ts           # CLI subprocess lifecycle & activity timeout
├── session/
│   └── manager.ts           # Conversation → CLI session mapping
├── server/
│   ├── routes.ts             # SSE streaming, progress notifications
│   ├── index.ts              # Express server setup
│   └── standalone.ts         # Entry point
└── types/
    ├── openai.ts             # OpenAI API type definitions
    └── claude-cli.ts         # CLI stream-json event types

Security

• No shell injection — Uses Node.js spawn(), not exec()
• No stored credentials — Authentication handled by Claude CLI's OS keychain
• No hardcoded secrets — All sensitive config via environment variables or external config files
• Local only by default — Binds to 127.0.0.1, not exposed to network

Tips

• Don't run heartbeat/cron jobs through Opus — Fixed-interval requests look like bot traffic. Use lightweight models for scheduled tasks.
• Stay within your weekly token limits — The proxy doesn't circumvent any usage caps. If you rarely hit your Claude Code weekly limit, you have plenty of headroom.

License

MIT

Credits

• Initial codebase based on atalovesyou/claude-max-api-proxy
• Session management, streaming, and OpenClaw integration built with Claude Code