mcp-neo4j-memory
Overview
The mcp-neo4j-memory server is a Model Context Protocol (MCP) server that provides AI assistants with persistent, graph-based memory backed by Neo4j. It allows AI-driven workflows to store, retrieve, and reason about entities, relationships, and interactions over time — turning long-term memory into a structured knowledge graph rather than a flat vector store.
This server is especially useful for agent memory, personalization, long-running assistants, and workflows where understanding relationships and history matters.
Transport
stdio
Tools
Key Capabilities
- Persistent agent memory — Retain information across sessions in a durable graph store.
- Relationship-aware context — Model connections between people, concepts, events, and actions.
- Temporal and evolving knowledge — Update and refine memory as new information is learned.
- Explainable retrieval — Retrieve memory based on explicit relationships, not just similarity scores.
- Knowledge graph workflows — Support reasoning that depends on structure, hierarchy, and connectivity.
How It Works
The mcp-neo4j-memory server runs as an MCP service connected to a Neo4j database and exposes memory operations over the MCP protocol. AI clients communicate with the server to persist new information or retrieve relevant context as part of broader reasoning workflows.
The server stores memory as nodes and relationships in Neo4j, enabling rich graph queries to identify related concepts, past interactions, or relevant entities. Results are returned in structured formats that AI assistants can reason over directly and combine with other MCP-provided context.
By representing memory as a graph rather than a flat store, the server enables AI workflows that can understand connections, track evolving relationships, and retrieve context with clear provenance — supporting more coherent, personalized, and explainable AI behavior over time.