A
licenseNot graded
qualityB
maintenanceExposes read-only long-term memory tools for LLM coding agents, letting them recall durable facts and decisions via fused vector and BM25 search, query decision histories with their supersession chains, and fetch a token-budgeted, role-filtered context payload at session start. Writes are handled elsewhere, so the server only surfaces curated knowledge scoped per project plus a shared global scope.
MIT