MCP server bridging Claude Code to local llama.cpp. Run local LLMs alongside Claude for experimentation, testing, and cost-effective inference.