shunt

Shunt is a transparent proxy that intercepts Claude Code requests and routes them to alternative LLM providers based on model ID, while passing everything else through to Anthropic unchanged. It implements the official Claude Code LLM gateway protocol with support for OpenAI, ChatGPT/Codex, xAI Grok, Cursor, Kimi, DeepSeek, GLM, MiniMax, and any Anthropic-compatible backend. Routing decisions are per-model and driven by request metadata rather than prompt inspection. Built with Claude Code and designed for developers who want to divert specific model inference to other providers without disrupting Claude Code's tools and capabilities.