OpenMemory FastMCP Server with Tailscale Integration

A FastMCP-based implementation of the OpenMemory server that replaces SSE (Server-Sent Events) with HTTP streaming transport, deployed as a Docker container with Tailscale network integration.

Overview

This project reimplements the mem0 OpenMemory MCP server using FastMCP's HTTP streaming transport, providing:

• HTTP Streaming: More reliable than SSE with better browser and proxy support
• Tailscale Integration: Secure remote access without port forwarding
• Docker Deployment: Easy containerized deployment with all dependencies
• Full mem0 Compatibility: All memory management features preserved

Features

• ✅ Add memories with automatic categorization
• ✅ Search memories using vector similarity
• ✅ List all user memories with permission filtering
• ✅ Delete memories with audit logging
• ✅ Multi-tenancy support (users and apps)
• ✅ Qdrant vector database integration
• ✅ Support for OpenAI, Anthropic, and Ollama models

Architecture

┌─────────────────┐     ┌──────────────────┐
│   MCP Client    │────▶│  Tailscale Net   │
└─────────────────┘     └──────────────────┘
                               │
                               ▼
                    ┌─────────────────────┐
                    │  OpenMemory Server  │
                    │   (FastMCP HTTP)    │
                    └─────────────────────┘
                               │
                    ┌──────────┴──────────┐
                    ▼                     ▼
            ┌──────────────┐      ┌──────────────┐
            │   Qdrant     │      │   SQLite     │
            │ Vector Store │      │   Database   │
            └──────────────┘      └──────────────┘

Quick Start

Prerequisites

• Docker and Docker Compose
• Python 3.11+
• Tailscale account (for remote access)
• OpenAI API key (or Anthropic API key)

Installation

1. Clone the repository:

git clone https://github.com/artur-nohup/openmemory-fastmcp.git
cd openmemory-fastmcp

2. Create environment file:

cp .env.example .env
# Edit .env with your API keys and Tailscale auth key

3. Deploy the server:

./scripts/deploy.sh

Connection Methods

Local Access

from fastmcp import Client

async with Client("http://localhost:8765/mcp") as client:
    await client.call_tool("add_memories", {
        "text": "I prefer dark mode",
        "user_id": "john_doe",
        "client_name": "my_app"
    })

Tailscale Access

# From anywhere in your Tailscale network
async with Client("http://openmemory-server:8765/mcp") as client:
    # Same API as local access
    ...

MCP Tools

add_memories

Add new memories to the system.

Arguments:

• text (str): The text to remember
• user_id (str, optional): User identifier
• client_name (str, optional): Application identifier

search_memory

Search through stored memories using semantic similarity.

list_memories

List all memories for a user.

delete_all_memories

Delete all accessible memories for a user.

Testing

Run the test suite:

python testing/test_comprehensive.py

Monitoring

# View all services
docker-compose ps

# View logs
docker logs openmemory-fastmcp

License

This project is based on the mem0 OpenMemory implementation.

Acknowledgments

• mem0ai for the original OpenMemory implementation
• FastMCP for the improved MCP framework
• Tailscale for secure networking

🤖 Generated with Claude Code