The travel adapter for Claude Code

What it does

Claude Code Router is middleware that intercepts API calls from Anthropic’s CLI coding tool and forwards them to other LLM providers. You keep the same chat interface and agent features, but the actual model can be DeepSeek, Gemini, a local Ollama instance, or anything else with an OpenAI-compatible endpoint. The router translates requests and responses so Claude Code remains none the wiser.

The interesting bit

Instead of forking Claude Code, the activate command injects environment variables that trick the native claude binary into talking to the router transparently. It is essentially a polite man-in-the-middle: Claude Code thinks it is calling Anthropic, but it is really calling your configured switchboard.

Key highlights

• Routes to OpenRouter, DeepSeek, Ollama, Gemini, Volcengine, SiliconFlow, and others.
• Transformer plugins adapt provider-specific request and response formats automatically.
• Switch models mid-session with the /model command, or route by task type (background, thinking, long context).
• Preset system exports configurations with secrets sanitized into reusable templates.
• Built-in non-interactive mode for GitHub Actions and headless CI runs.

Caveats

• Without an explicit API key configured, the server is forcibly restricted to 127.0.0.1, which suggests the authors do not trust the internet with your proxy.
• The configuration file is expansive and JSON-heavy; the project provides a web UI and interactive CLI specifically because manual editing is unwieldy.

Verdict

Grab it if you want Claude Code’s workflow with a different model bill. Ignore it if you are content with Anthropic’s native API and do not need a configurable middle layer.