Claude Code
Claude Code talks the Anthropic Messages API shape (/v1/messages), while Draven AI's public API speaks the OpenAI Chat Completions shape (/v1/chat/completions) — same as the note on the Codex CLI page. There's no way to point Claude Code straight at a Chat Completions endpoint, so you run a small local proxy that translates between the two shapes.
Installation
npm install -g @anthropic-ai/claude-code
Verify with claude --version.
Bridging the API shape
claude-code-router is an open-source local proxy built for exactly this: it accepts Claude Code's Anthropic-shaped requests and forwards them to any OpenAI-compatible provider, translating the request/response on the fly.
npm install -g @musistudio/claude-code-router
Configuration
Point the router at Draven AI in ~/.claude-code-router/config.json:
{
"Providers": [
{
"name": "dravenai",
"api_base_url": "https://api.dravenai.lat/v1/chat/completions",
"api_key": "sk-drv-your-api-key",
"models": ["your-model-id"]
}
],
"Router": {
"default": "dravenai,your-model-id"
}
}
Pick your-model-id from the Models page — a strong reasoning/agentic model is a good default since Claude Code leans on tool-calling for file edits and shell commands.
Running Claude Code
Start Claude Code through the router instead of calling it directly:
ccr code
That launches the router in the background (default http://127.0.0.1:3456) and runs claude with ANTHROPIC_BASE_URL already pointed at it, so no manual env var juggling is needed.