Best MCP Servers for Memory for AI Agents (2026)
Long-term memory is what turns a stateless chat into an assistant that knows your project. Cognee and Hindsight build knowledge-graph memory, the MCP Memory Service and Mem0 add persistent recall to any client, and lighter options like Shodh, Mengram, and Prism keep memory local.
This page collects the 13 MCP servers in our directory that genuinely cover memory for ai agents — together they have 32,816 GitHub stars, and 0 of them run as hosted services you can connect to without installing anything. The highest-rated options are topoteretes/cognee, vectorize-io/hindsight, doobidoo/mcp-memory-service.
How to choose
Choose Cognee or Hindsight for graph-based memory over documents, MCP Memory Service or Mem0 for a general-purpose memory layer, and a local option like Shodh if everything must stay on your machine.
topoteretes/cognee
by topoteretes
📇 🏠 - Memory manager for AI apps and Agents using various graph and vector stores and allowing ingestion from 30+ data sources
npx -y topoteretes-cogneevectorize-io/hindsight
by vectorize-io
Hindsight is an agent memory system built so agents learn over time, rather than only recalling conversation history. It organizes long-term memory into banks with three core operations, retain, recall, and reflect, plus observations and mental models, and it can be added to an agent through integrations that include an MCP server, client SDKs for Python and TypeScript, or an embedded Python mode with no server required.
npx -y skillsdoobidoo/mcp-memory-service
by doobidoo
📇 🏠 - Universal memory service providing semantic search, persistent storage, and autonomous memory consolidation
uvx CIagentic-mcp-tools/memora
by agentic-mcp-tools
🐍 🏠 ☁️ - Persistent memory with knowledge graph visualization, semantic/hybrid search, cloud sync (S3/R2), and cross-session context management.
npx -y wranglerCheMiguel23/MemoryMesh
by CheMiguel23
📇 🏠 - Enhanced graph-based memory with a focus on AI role-play and story generation
npx -y memorymeshl33tdawg/sage
by l33tdawg
](https://glama.ai/mcp/servers/l33tdawg/s-age) 🏎️ 🏠 🍎 🪟 🐧 - Institutional memory for AI agents with real BFT consensus. 4 application validators vote on every memory before it's committed — no mo
npx -y sagevarun29ankuS/shodh-memory
by varun29ankuS
Shodh-Memory is persistent cognitive memory for AI agents and robots that runs as a single binary with no LLM in the loop, no API keys, and no external databases, so it works fully offline. Storing and recalling memories uses local algorithms only: on-device embeddings, a cross-encoder reranker, and local entity and relation extraction that builds a typed knowledge graph; memories that are used often become easier to recall (Hebbian learning) while irrelevant ones decay. It connects to agents over MCP/HTTP and to robots over Zenoh/ROS2.
npx -y -y @shodh/memory-mcpalibaizhanov/mengram
by alibaizhanov
](https://glama.ai/mcp/servers/@alibaizhanov/mengram) 🐍 ☁️ 🏠 🍎 🪟 🐧 - Human-like memory layer for AI agents with semantic, episodic, and procedural memory. Claude Code hooks (auto-save, auto-recal
npx -y alibaizhanov-mengramdcostenco/prism-mcp
by dcostenco
](https://glama.ai/mcp/servers/dcostenco/BCBA) 📇 🏠 🍎 🪟 🐧 - Zero-config persistent memory for AI agents with local SQLite. Mind Palace web dashboard, time travel (rewind/replay sessions), agent te
npx -y -y prism-mcp-serveromega-memory/omega-memory
by omega-memory
🐍 🏠 🍎 🪟 🐧 - Persistent memory for AI coding agents with semantic search, auto-capture, cross-session learning, and intelligent forgetting. 12 MCP tools, local-first.
npx -y omega-memory-omega-memoryelvismdev/mem0-mcp-selfhosted
by elvismdev
](https://glama.ai/mcp/servers/elvismdev/mem0-mcp-selfhosted) 🐍 🏠 🍎 🪟 🐧 - Self-hosted mem0 MCP server for Claude Code with Qdrant vector search, Neo4j knowledge graph, and Ollama embeddings. Zero
uvx --fromaitytech/agentkits-memory
by aitytech
](https://glama.ai/mcp/servers/@aitytech/agentkits-memory) 📇 🏠 🍎 🪟 🐧 - Persistent memory for AI coding assistants with hybrid search (FTS5 + vector embeddings), session tracking, automatic contex
npx -y @aitytech/agentkits-memoryMemory
by Anthropic
Persistent memory storage for Claude. Save and retrieve information across conversations using a knowledge graph.
npx @modelcontextprotocol/server-memoryFrequently asked questions
What is the most popular MCP server for Memory for AI Agents?
topoteretes/cognee leads this list with 17,048 GitHub stars. 📇 🏠 - Memory manager for AI apps and Agents using various graph and vector stores and allowing ingestion from 30+ data sources
Are these Memory for AI Agents MCP servers free to use?
12 of the 13 servers here publish a license in the directory (Apache-2.0, MIT, NOASSERTION). 0 of them run as hosted services you can connect to remotely; the rest you install and run yourself. Hosted services may have their own pricing.
How do I choose an MCP server for Memory for AI Agents?
Choose Cognee or Hindsight for graph-based memory over documents, MCP Memory Service or Mem0 for a general-purpose memory layer, and a local option like Shodh if everything must stay on your machine.