
Cozo Memory
This is a local-first memory system that runs CozoDB embedded in Node.js, giving Claude persistent storage with hybrid search across vector embeddings, full-text, and graph traversals. It exposes entity CRUD, observation tracking, and three search modes: vector similarity via HNSW, multi-hop graph reasoning, and community-based summaries. The agentic retrieval layer auto-routes queries to the right strategy, and temporal validity lets you query any point in history. Embeddings run locally via ONNX (bge-m3 by default, swap to MiniLM if you're RAM-constrained). Includes LangChain and LlamaIndex adapters. Reach for this when you need structured, queryable memory that goes beyond append-only chat history, especially if you want graph relationships and time-travel without managing external databases.
- GitHub stars
- 1
- tools
- 5
- transport
- STDIO
- last updated
- Feb 28, 2026
Tools
5 toolsPublic tool metadata for what this MCP can expose to an agent.
mutate_memoryWrite operations: create_entity, update_entity, delete_entity, add_observation, create_relation, transactions, sessions, tasks, update_observation, batch_delete, manage_tags, batch.
Write operations: create_entity, update_entity, delete_entity, add_observation, create_relation, transactions, sessions, tasks, update_observation, batch_delete, manage_tags, batch.
No parameter schema in public metadata yet.
query_memoryRead operations: search, advancedSearch, context, graph_rag, graph_walking, agentic_search, adaptive_retrieval, list_entities, get_entity_detail, get_session_context, list_sessions.
Read operations: search, advancedSearch, context, graph_rag, graph_walking, agentic_search, adaptive_retrieval, list_entities, get_entity_detail, get_session_context, list_sessions.
No parameter schema in public metadata yet.
analyze_graphGraph analysis: explore, communities, pagerank, betweenness, hits, shortest_path, semantic_walk.
Graph analysis: explore, communities, pagerank, betweenness, hits, shortest_path, semantic_walk.
No parameter schema in public metadata yet.
manage_systemMaintenance: health, metrics, stats, export, import, cleanup, defrag, reflect, snapshots.
Maintenance: health, metrics, stats, export, import, cleanup, defrag, reflect, snapshots.
No parameter schema in public metadata yet.
edit_user_profileEdit global user profile with preferences and work style.
Edit global user profile with preferences and work style.
No parameter schema in public metadata yet.
Sponsors
Readme
CozoDB Memory MCP Server
Why Cozo Memory?
LLMs have short-term memory limits. Standard RAG retrieves documents but can't connect facts across time. Cozo Memory gives your AI agent persistent, structured memory – it remembers past conversations, infers relationships, detects contradictions, and explores its knowledge graph – fully on your machine, with optional local LLM integration via Ollama for intelligent actions (cleanup, reflection, summarization, agentic routing).Most memory stacks combine separate databases: SQLite for facts, Chroma for vector search, NetworkX for graphs. CozoDB replaces all of that with one embedded engine: relational, graph, vector, and full-text search in a single query language, one file, zero sync lag.
Details
- Repository
- github.com/tobs-code/cozo-memory
- Categories
- AI & LLM Tools, Search & Web Crawling
- Package
cozo-memory- Transport
- STDIO
- Registry
- MCP registry, active