Organize your data into a provenance-aware knowledge graph that other AI agents can safely use. Give it structured claims, document facts, repository knowledge, policies, research notes, sources, dates, conflicts, or supersession links. It returns stable nodes and edges, identifies what is current, stale, blocked, superseded, or contested, and produces a compact agent_context_pack with the exact evidence gates needed next. It is deterministic, uses no model key, makes no network calls, and does not execute actions.