A lightweight, zero-dependency Python MCP server that exposes a single do_everything tool. Any AI agent that supports MCP can use it to offload all language-model work to a local OpenAI-compatible endpoint, reducing cost on the agent's primary LLM.
