engram¶
a cognitive memory system that actually remembers things.
starts with one sqlite file for local installs, can switch to postgres for multi-client deployments, and keeps hybrid retrieval that fuses five signals, memory layers that model how brains actually work, and a neural visualization that shows the whole thing firing in real time.
98.1% R@5 on LongMemEval (ICLR 2025) — highest published score, beating MemPalace (96.6%), Emergence AI (86%), and every other memory system benchmarked.
what it does¶
- hybrid retrieval — HNSW dense + BM25 + entity graph BFS + Hopfield associative pattern completion, fused with intent-weighted reciprocal rank fusion, cross-encoder reranking, deep MLP reranking
- memory layers — working, episodic, semantic, procedural, codebase. memories promote upward when useful, decay if unused
- entity graph — extracts people, tools, projects from every memory. multi-hop traversal via recursive SQL CTEs
- session continuity — active MCP sessions now maintain a structured handoff packet with recent work, decisions, open loops, touched entities, and search history. new sessions can call
resume_contextinstead of reconstructing state from raw logs - memory intelligence — query briefing, retrieval comparison, and recent hotspot surfacing help you see what matters without manually stitching together raw results
- 83 MCP tools — plugs into Claude Code or any MCP client. docker-ready
- multi-backend embeddings — local (MLX/CPU), Voyage AI, OpenAI, Google Gemini. auto-detects from model name
- web dashboard — neural map, search, continuity, intelligence, analytics, cognition, drift detection, and more in one surface
install¶
or from source:
quick links¶
- Installation — full setup guide
- Quick Start — first steps
- Postgres Migration — move an existing sqlite install without losing data
- MCP Tools Reference — all 83 tools
- Architecture — how it works inside
- Benchmarks — LongMemEval results
supported by¶
| system | what |
|---|---|
| Claude Code | native MCP integration |
| Any MCP client | 83-tool JSON-RPC server |
| REST API | 57 HTTP endpoints |
| CLI | 16 commands |
| Docker | single container deployment |
| Python API | direct library usage |