A shared, queryable memory database for AI assistants. Give your agents persistent context with keyword and semantic search, journaling, and real-time consolidation.
Hybrid retrieval combining keyword matching with vector similarity. Falls back to semantic search automatically when keyword results are sparse.
Ingest raw context from chat sessions. An automated processor refines entries with LLM-generated keywords and summaries using Bedrock.
Native Model Context Protocol server. Claude Code, Cursor, and other MCP-compatible tools can search and write to memory directly.
AI-powered summarization layer. Agents send a query, get back a concise summary synthesized from relevant memories by Claude.
Organize memories by team, project, or domain. Users map to shards with role-based permissions. Cross-shard search supported.
Automatic graph-based deduplication. Related entries are merged nightly into consolidated summaries, keeping the knowledge base lean.
The installer sets up memory-context-server (MCP), memory-admin (CLI), and memory-web (dashboard). Requires SSO authentication.
Signed download links for your platform. Links expire after 1 hour.