Repolode
Submit
doobidoo

doobidoo/mcp-memory-service

Self-hosted persistent memory service for AI agents, exposed via REST API and MCP so frameworks and assistants like Claude can store and retrieve shared context.

Open-source alternative to:Mem0

View on GitHub
Stars
2,091
Forks
395
Open issues
49
Last commit
17 hours ago

What is it?

mcp-memory-service is an open-source memory backend for AI agents that keeps decisions, notes and project context across sessions. It runs as one self-hosted service that offers a REST API, an MCP server, OAuth, a CLI and a web dashboard, so agents built with LangGraph, CrewAI or AutoGen, plain HTTP clients, and MCP clients such as Claude Desktop and Claude Code can all read and write the same memory store.

Beyond basic storage, it combines vector search with BM25 keyword search and supports tag and time-range filtering. Memories can be linked into a knowledge graph with typed relationships, consolidated automatically over time, and imported from PDF, text and Markdown files or from mem0 exports. Storage options include local SQLite, Cloudflare, Milvus, and a Hybrid backend that syncs local reads with a remote secondary. Embeddings run locally via ONNX.

The server binds to localhost by default, and exposing it on a network requires explicit configuration and authentication.

Key features

  • 01REST API and MCP server share one memory store across agents and clients.
  • 02Hybrid search combines vector similarity with BM25 keyword matching and tag filters.
  • 03Knowledge graph stores typed relationships between memories.
  • 04Automatic consolidation clusters and compresses older memories.
  • 05Local ONNX embeddings with SQLite storage keep core operations on your own infrastructure.
  • 06Imports PDF, text, Markdown and mem0 exports; includes a web dashboard and CLI.

Use cases

  • Sharing project decisions and architecture notes between LangGraph or CrewAI agents.
  • Giving Claude Desktop or Claude Code persistent memory across chat sessions.
  • Running a self-hosted memory service for a small team's AI tools.
  • Migrating memories from mem0 into a self-hosted store.

Quick start

  1. 1

    Install the package from PyPI

    pip install mcp-memory-service
  2. 2

    Add the memory server to your AI client config (for example Claude Desktop)

  3. 3

    Start the REST API server on localhost

    MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http

Things to know

  • The quick HTTP setup allows anonymous access, so secure it before exposing it to a network.
  • Binding to a network address requires authentication, firewall rules, or a VPN or TLS termination.
  • Optional LLM features such as search summaries need a configured LLM provider.

In collections

Long-term memory for AI agentsOpen-source AI agent frameworks & appsMCP serversOpen-source personal AI agents

Topics

#agent-memory#ai-agents#mcp-server#vector-database#knowledge-graph#semantic-search#rest-api#personal-agent#agentic-ai#autogen

FAQ

What is mcp-memory-service?
Self-hosted persistent memory service for AI agents, exposed via REST API and MCP so frameworks and assistants like Claude can store and retrieve shared context. mcp-memory-service is an open-source memory backend for AI agents that keeps decisions, notes and project context across sessions. It runs as one self-hosted service that offers a REST API, an MCP server, OAuth, a CLI and a web dashboard, so agents built with LangGraph, CrewAI or AutoGen, plain HTTP clients, and MCP clients such as Claude Desktop and Claude Code can all read and write the same memory store.
Is mcp-memory-service free to use? What license does it have?
mcp-memory-service is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is mcp-memory-service still maintained?
Actively maintained. The last commit to mcp-memory-service was on Nov 6, 2026.
What can I use mcp-memory-service for?
Sharing project decisions and architecture notes between LangGraph or CrewAI agents. Giving Claude Desktop or Claude Code persistent memory across chat sessions. Running a self-hosted memory service for a small team's AI tools. Migrating memories from mem0 into a self-hosted store.
How do I install mcp-memory-service?
Install the package from PyPI: pip install mcp-memory-service Add the memory server to your AI client config (for example Claude Desktop) Start the REST API server on localhost: MCP_ALLOW_ANONYMOUS_ACCESS=true memory server --http
Where can I get mcp-memory-service?
The source code and installation instructions are on GitHub: https://github.com/doobidoo/mcp-memory-service

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects