cmappy

OpenAI-compatible API proxy for Claude Max subscribers. Wraps the Claude Code CLI as a subprocess and exposes a standard /v1/chat/completions endpoint — zero API keys, zero extra cost.

Your App (any OpenAI-compatible client)
         ↓
    HTTP Request (OpenAI format)
         ↓
   cmappy (FastAPI, port 3456)
         ↓
   Claude Code CLI (subprocess)
         ↓
   OAuth Token (from Max subscription)
         ↓
   Anthropic API
         ↓
   Response → OpenAI format → Your App

Prerequisites

1. Claude Max subscription ($100 or $200/month)
2. Claude Code CLI installed and authenticated:

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

3. Python 3.10+ and uv

Quick start

# install (pick one)
uv tool install claude-max-proxy-py   # uv
pip install claude-max-proxy-py       # pip

# run
cmappy

The server starts on http://127.0.0.1:3456. Point any OpenAI-compatible client at it.

Install

# from PyPI
uv tool install claude-max-proxy-py   # uv (recommended)
pip install claude-max-proxy-py       # pip

# from source
git clone https://github.com/p-sumann/claude-max-proxy-py.git
cd claude-max-proxy
uv tool install .   # or: pip install .

Uninstall

uv tool uninstall claude-max-proxy-py   # uv
pip uninstall claude-max-proxy-py       # pip

Usage

# default
cmappy

# custom port
cmappy --port 8080

# bind to all interfaces
cmappy --host 0.0.0.0

# skip auth check on startup
cmappy --skip-auth-check

# dev mode (auto-reload)
cmappy --reload

API

Models

curl

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

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}'

OpenAI Python SDK

from openai import OpenAI

client = OpenAI(base_url="http://localhost:3456/v1", api_key="not-needed")
response = client.chat.completions.create(
    model="claude-sonnet-4",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

How it works

The proxy spawns the claude CLI with these flags:

claude --print --output-format stream-json --verbose \
  --include-partial-messages --model <alias> \
  --no-session-persistence "<prompt>"

It reads stdout as NDJSON and classifies each line into events: content_block_delta (text chunks for SSE), assistant (model name), and result (final response with usage stats). These are converted on-the-fly into OpenAI-format responses.

Development

git clone https://github.com/p-sumann/claude-max-proxy-py.git
cd claude-max-proxy

# install with dev deps
uv sync --dev

# run without installing
uv run cmappy

# run tests
uv run pytest

# lint
uv run ruff check .

Contributing

Contributions are welcome! Please read the contributing guide before opening a PR.

Security

If you discover a security vulnerability, please see our security policy. Do not open a public issue for security reports.

License

MIT