MCPFast / Tools / Persistent MCP memory server for AI agents: SQLite + FTS5 + vector hybrid search

GitHubMCP★★★★☆

Persistent MCP memory server for AI agents: SQLite + FTS5 + vector hybrid search

An open-source MCP server for persistent AI agent memory, combining SQLite, FTS5, and hybrid vector search (RRF) without the cloud.

View on GitHub→

Persistent MCP Memory Server for AI Agents

This tool provides a self-hosted, persistent memory server for AI agents, leveraging the power of SQLite with FTS5 for full-text search and vector embeddings for semantic recall. Designed for developers building sophisticated AI agents, it offers a robust and efficient solution for managing and retrieving agent memories without relying on external cloud services. This MCP server enables your agents to build a comprehensive and searchable history of their interactions and learned information, crucial for complex task execution and long-term planning.

What it Does

The Persistent MCP Memory Server acts as a central repository for an AI agent's memories. It stores data in a structured SQLite database, enhanced with FTS5 for fast and accurate keyword-based searches. Simultaneously, it integrates vector embeddings, allowing for semantic similarity searches. This hybrid approach ensures that agents can retrieve relevant information based on both exact matches and conceptual understanding, significantly improving their ability to recall past experiences and knowledge. The entire system runs locally, providing full control over data and infrastructure.

Key Features

SQLite Backend: Reliable and ACID-compliant data storage.
FTS5 Integration: High-performance full-text search capabilities for keyword retrieval.
Vector Hybrid Search (RRF): Combines keyword and semantic search results for more accurate recall.
Self-Hosted: No cloud dependencies, ensuring data privacy and control.
MCP Protocol: Seamless integration with existing MCP agent frameworks.
Open-Source: Transparent and customizable for developer needs.

Who it's For

This tool is specifically designed for AI developers and researchers building agents that require persistent memory. It is ideal for:

AI Engineers: Implementing memory systems for complex AI agents.
Researchers: Experimenting with long-term memory architectures for AI.
Developers: Creating autonomous agents that learn and adapt over time.
Teams: Requiring a self-hosted, secure, and scalable memory solution for their AI projects.