Our mission is to improve education by building AI that matters.

An LLM agent is only as current as the text it reads. GROOM keeps the corpus current by making each consultation the trigger to refresh it, and gates every edit behind a checkpoint. Format-agnostic, and it speaks the Open Knowledge Format.

Agent quality lives below the model: the loop, the context engine, the tool surface, the safety stack, and the evaluator. A survey of harness engineering, introducing GROOM.