Mem0 MCP Server

# Mem0 MCP Server

A Model Context Protocol (MCP) server that provides memory capabilities using Mem0, backed by local Postgres (graph store) and Qdrant (vector store).

Features

• Tools: add_memory, add_fact, search_memories, get_all_memories, database_history

• Storage: Local Postgres (Graph) & Qdrant (Vector)

• Transport: Supports both stdio (local) and sse (remote/docker)

• Deployment: Ready for Docker Compose and GitHub Actions

Prerequisites

• Docker & Docker Compose

• Python 3.12+ (for local development)

Quick Start (Docker)

The easiest way to run the full stack (Database + MCP Server) is via Docker Compose:

1. Configure Environment

   cp .env.example .env
   # Edit .env to add your OPENAI_API_KEY if needed

2. Start Services

   # Starts Postgres, Qdrant, and the MCP Server (SSE mode on port 8000)
   docker-compose -f docker-compose.mem0.yml up -d --build

3. Connect Client

   • The server runs in SSE mode on port 8000.

   • SSE Endpoint: http://localhost:8000/sse

Local Development (Python)

If you want to run the server code locally while keeping databases in Docker:

1. Start Databases Only

   docker-compose -f docker-compose.mem0.yml up -d db qdrant

2. Install Dependencies

   python -m venv .venv
   source .venv/bin/activate  # or .venv\Scripts\activate on Windows
   pip install .

3. Run Server

   • Stdio Mode (Default, for use with local MCP clients like Claude Desktop):

     python src/main.py

   • SSE Mode (Network accessible):

     export MCP_TRANSPORT=sse
     python src/main.py
     # Server lists on 0.0.0.0:8000

Deployment

The repository includes a GitHub Actions workflow (.github/workflows/deploy.yml) to deploy to a remote server (e.g., 192.168.254.202).

• DB Deploy: Deploys db and qdrant services.

• App Deploy: Deploys the mcp-server container in SSE mode, mapping port 8000.

Configuration

Tools

• add_memory(content, user_id, metadata): Stores a user message.

• add_fact(content, user_id, metadata): Stores a specific fact (tagged with type: fact).

• search_memories(query, user_id, limit): Semantic search for relevant memories.

• get_all_memories(user_id, limit): Retrieves all stored memories for a specific user.