# Agent memory MCP servers

MCP servers that give an agent long-term memory: store facts, recall them by meaning and build knowledge graphs.

## Before you choose

Remote memory services keep what the agent stores on their side; local servers keep it in a file or database on your machine.

Whatever the agent remembers may include personal data, so check where each server stores it and how you delete it.

## Servers (20)

What search for "memory" returns, without listings scored poor, published under many names from one repository, or with nothing to connect to.

- [WorkingMemory](https://protogrid.dev/servers/ai.workingmemory/memory.md) `ai.workingmemory/memory`: class R0, quality 97 (strong), trust 75. Persistent personal memory for AI assistants — save, search, and recall across every MCP client.
- [Cairn.ink Memory](https://protogrid.dev/servers/ink.cairn/memory.md) `ink.cairn/memory`: class R2, quality 83 (good), trust 95. Cross-session memory for AI agents with a Source Receipt for every memory, over MCP.
- [AiAkiv Memory](https://protogrid.dev/servers/com.aiakiv/memory.md) `com.aiakiv/memory`: class R2, quality 100 (strong), trust 100. Shared long-term memory for AI agents: save and recall context as a searchable knowledge graph.
- [Korely Memory](https://protogrid.dev/servers/ai.korely/memory.md) `ai.korely/memory`: class R2, quality 83 (good), trust 96. Memory for AI agents that knows what is still true. Typed bi-temporal facts, EU-hosted.
- [Aziel Runtime](https://protogrid.dev/servers/io.github.AzielEliab/aziel-runtime.md) `io.github.AzielEliab/aziel-runtime`: class R0, quality 80 (good), trust 96. Governed MCP: agent audit, provenance, deterministic checks, and receipt-backed FragGate execution.
- [ai.mitosislabs/mitosis](https://protogrid.dev/servers/ai.mitosislabs/mitosis.md) `ai.mitosislabs/mitosis`: class R0, quality 83 (good), trust 88. Mitosis agent-memory platform: pricing, docs search, platform status, agent skills. No auth.
- [MeshMarket](https://protogrid.dev/servers/ai.meshtool/meshmarket.md) `ai.meshtool/meshmarket`: class R0, quality 81 (new), trust 96. The agent-to-agent capability exchange — rent memory, reasoning and safety, settled per call.
- [Mnemoverse Memory](https://protogrid.dev/servers/io.github.mnemoverse/mcp-memory-server.md) `io.github.mnemoverse/mcp-memory-server`: class R2, quality 83 (good), trust 97. Hosted agent memory that learns from outcomes, with shared rooms. No key: mcp.mnemoverse.com/mcp
- [Kortexio Memory](https://protogrid.dev/servers/io.kortexio/memory.md) `io.kortexio/memory`: class unknown, quality – (not scored), trust 96. Hosted ambient memory for AI agents. Retrieve context and save notes via MCP tools with OAuth.
- [Worldku Memory](https://protogrid.dev/servers/com.worldku/memory.md) `com.worldku/memory`: class R2, quality 100 (strong), trust 92. Save AI conversations into your Worldku memory and recall them from any MCP client.
- [MCP Memory](https://protogrid.dev/servers/io.github.GermaniU/mcp-memory.md) `io.github.GermaniU/mcp-memory`: class L0, quality – (new), trust 90. Local persistent memory for AI agents: SQLite + FTS5 (BM25) lexical search, zero external infra.
- [Mem0 Cloud](https://protogrid.dev/servers/com.saastemly/mem0-cloud.md) `com.saastemly/mem0-cloud`: class R0, quality 79 (good), trust 82. A hosted agent-memory API shaped like mem0's REST API (an alternative to running mem0 yourself):...
- [Gnosis Memory](https://protogrid.dev/servers/com.gnosismemory/memory.md) `com.gnosismemory/memory`: class R2, quality 83 (good), trust 88. Persistent memory and vector search for AI agents. Hosted, OAuth-protected via Google sign-in.
- [com.humaux/memory](https://protogrid.dev/servers/com.humaux/memory.md) `com.humaux/memory`: class R2, quality 83 (good), trust 88. Persistent long-term memory for AI agents: semantic search, knowledge graph, and task canvas.
- [core.blue](https://protogrid.dev/servers/blue.core/memory.md) `blue.core/memory`: class R0, quality 86 (good), trust 83. Private, self-describing memory for agents: one house per customer, on a machine of its own.
- [Valet](https://protogrid.dev/servers/io.github.valetdotdev/valet.md) `io.github.valetdotdev/valet`: class R0, quality 89 (good), trust 77. Get share links, publish and manage websites, artifacts and agents. No account needed.
- [MemoryRouter](https://protogrid.dev/servers/ai.memoryrouter/memoryrouter.md) `ai.memoryrouter/memoryrouter`: class R0, quality 94 (strong), trust 79. Persistent AI memory shared across Claude, ChatGPT, coding agents, and compatible MCP clients.
- [ASR Memory (에사 AI) — Shared memory for AI tools](https://protogrid.dev/servers/com.asrmemory/asr.md) `com.asrmemory/asr`: class R2, quality 83 (good), trust 96. Shared memory for AI tools: save once, recall word for word from Claude, ChatGPT, Codex or Gemini.
- [Kythene](https://protogrid.dev/servers/com.kythene/kythene.md) `com.kythene/kythene`: class R0, quality 89 (good), trust 97. Your AI's work, reviewed by your team, remembered - shared artifacts with review, plus team memory.
- [ctxstore — memory that follows you across models](https://protogrid.dev/servers/ai.ctxstore/ctxstore-mcp.md) `ai.ctxstore/ctxstore-mcp`: class R0, quality 85 (good), trust 96. Persistent memory for AI agents: keyed facts, a wake bundle each session, one memory in every model.

All use cases: https://protogrid.dev/mcp.md