Anclum Submit

Best MCP servers for persistent memory

8 MCP servers that let an agent remember between sessions — a local graph file, a searchable service, or a hosted store. Where the data lives decides most of the rest.

How to choose

Memory is the one category here where "which one is best" is mostly a question about your own answer to four things.

Our call: start with a local graph server scoped to one project, keep its file inside that project (or gitignored on purpose, not by accident), and move to a shared service only when a second agent or a second machine needs the same memory.

Compare

ToolWhat it remembersInstallStars
modelcontextprotocol/server-memory Site ↗Knowledge graph-based persistent memory for maintaining context—91,352
vectorize-io/hindsight Site ↗Hindsight: long-term memory for AI agents modeled on human memory—51,200
topoteretes/cognee Site ↗Memory manager for AI apps and agents using graph and vector stores, with ingestion from m—32,487
cjo4m06/mcp-shrimp-task-manager Site ↗Task management for coding agents with task memory, self-reflection and dependency managem—2,146
doobidoo/mcp-memory-service Site ↗Memory service with semantic search, persistent storage and autonomous memory consolidatio—2,021
neo4j-contrib/mcp-neo4j Site ↗Neo4j integration: run queries, use Neo4j as knowledge graph memory and manage Neo4j Aura —988
agentic-mcp-tools/memora Site ↗Persistent memory with knowledge graph visualization, semantic/hybrid search, cloud sync (—742
CheMiguel23/MemoryMesh Site ↗Graph-based memory focused on AI role-play and story generation—354

Why these links: every card opens the entry's own page here first (what it does, how to install, how to verify the numbers); the tool's official site is one click further, and nothing here is a paid placement. Numbers come from public sources and are shown so you can check them.

Keep exploring: MCP directory · all fields · all AI tools · rankings