ProductivityOfficial
Memory MCP Server
Persistent knowledge graph — facts, entities, relations.
create_entitiesadd_observationssearch_nodes
The memory MCP server gives your AI client a persistent knowledge graph — entities, observations, and relations that survive across sessions. Unlike a chat history (lost when the conversation ends) or RAG (retrieved per-query from documents), memory accumulates facts as structured triples and lets the AI build up context over time about your projects, people, and decisions.
Install in your client
Copy and paste — no manual config. Replace placeholder values.
run in any terminal
claude mcp add memory npx -y @modelcontextprotocol/server-memoryPackage: @modelcontextprotocol/server-memory
Common use cases
- Have the AI remember decisions you've made about your codebase across sessions
- Build a personal knowledge base by adding observations from conversations
- Use as a structured store for project metadata that informs future prompts
Note
Memory is most useful when paired with a discipline of telling the AI when to write to memory ('save this'). Without that, it stays empty.