Persistent, maintainable, and testable expert context for AI agents. Open specification + reference implementation with eval-gated maintenance. Ground your agents with auditable citations, incremental refresh, and built-in quality controls.