A lightweight, Ollama-compatible inference server that answers questions through knowledge retrieval, structured reasoning chains, and MCP tool calling — with no model weights, no training, and no GPU required.

It scores 97% on a 35-question benchmark spanning graduate-level science, medicine, law, finance, and software engineering. It runs on a laptop CPU in under 1ms per query.

Fun fact: This was able to summarize the Lord of the Rings well and work on a variety of different problems.