Claude Code Bridge

The bridge lets routing keys reach Claude Code. It is a small service that runs on your server, next to your Claude Code sign-in, and exposes an OpenAI-compatible endpoint that ProxyLLM can route to.

Why the bridge runs on your server

Anthropic’s terms do not allow ProxyLLM to run Claude Code for you in our cloud, so we never host it, and your Claude sign-in never touches our servers. The bridge wraps Claude Code’s programmatic mode (claude -p) for testing purposes on infrastructure you own. ProxyLLM stores only connection metadata (the bridge URL and token, plus health status). You stay responsible for your cloud account, your Claude Code sign-in, and the provider’s terms.

Set it up

You need a small Linux server (any cloud works). On it, run:

npx proxyllm-bridge setup

The setup command walks you through everything: it signs you into your ProxyLLM account (or takes a pasted account token), installs Claude Code if needed, starts the bridge, and registers it to your account. When it finishes, the bridge shows up under Seats.

Don’t have a server? The dashboard flow can create one in your own cloud account (DigitalOcean, Hetzner, and others) with your authorization. It is still your server on your bill; ProxyLLM just bootstraps the bridge onto it.

Use it

Add a bridge lane to a routing key under Routing keys, then request the model bridge/claude-code:

curl https://api.proxyllm.ai/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bridge/claude-code",
    "messages": [{"role": "user", "content": "Say hello."}]
  }'

Health checks run automatically; a failing bridge falls through to the next lane on the key (how routing works).