Production-ready MCP memory service with zero database locking, hybrid backend (fast local reads + cloud synchronization) and intelligent memory search for AI assistants. Includes auto-configuration for multi-client access, local reads in 5ms with background synchronization through Cloudflare, Natural Memory Triggers with 85%+ accuracy, and team collaboration via OAuth 2.1.

Installation

PyPI

pip install mcp-memory-service

UV

uv pip install mcp-memory-service

From Source

git clone https://github.com/doobidoo/mcp-memory-service.git
cd mcp-memory-service && python install.py

Docker

docker-compose up -d

Smithery

npx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude

Configuration

Claude Desktop

{
  "mcpServers": {
    "memory": {
      "command": "memory",
      "args": ["server"],
      "env": {
        "MCP_MEMORY_STORAGE_BACKEND": "hybrid"
      }
    }
  }
}

Features

• Zero database locking with concurrent access
• Hybrid backend with fast local reads (5ms) and cloud synchronization
• Natural Memory Triggers with 85%+ accuracy
• Intelligent memory search and automatic context injection
• Team collaboration via OAuth 2.1
• Memory consolidation system with decay scoring inspired by dreams
• Multi-client support (Claude Desktop, VS Code, Cursor, Continue, and 13+ AI applications)
• SQLite-vec with ONNX embeddings for offline operation
• Background synchronization through Cloudflare
• Git-aware context integration

Environment Variables

Optional

• MCP_MEMORY_STORAGE_BACKEND - Storage backend type (hybrid, cloudflare, sqlite)

Notes

Supports Python 3.12+ (note on Python 3.13 compatibility for sqlite-vec). macOS users may need Homebrew Python for SQLite extension support. Initial setup includes automatic model download (~25MB). The visible memory embedding feature displays top-3 memories at session start with relevance scores.