
SQLite Memory MCP
by rmanov
Production-grade SQLite-backed memory stack for LLMs with concurrent safety and FTS5 search.
What it does
SQLite Memory MCP provides a persistent, high-performance memory layer for AI agents and LLMs. It replaces fragile JSONL-based memory with a robust SQLite database using Write-Ahead Logging (WAL) to allow multiple concurrent sessions without file-locking conflicts. It enables agents to maintain long-term knowledge graphs, track tasks, and sync state across different machines.
Tools
The server provides a comprehensive suite of 57 tools across several specialized servers:
sqlite_memory: Core knowledge-graph tools for creating entities and observations.sqlite_tasks: Structured task management including CRUD, digests, and priority bumping.sqlite_session: Session tracking and context recall for continuity across restarts.sqlite_bridge: Cross-machine synchronization via private git repositories.sqlite_collab: Collaborative knowledge sharing and public knowledge workflows.sqlite_entity: Tooling for linking tasks to entities and merging overlaps.sqlite_intel: Context assessment and enrichment tools for agent intelligence.
Installation
To install and add the unified server to Claude Desktop:
pip install sqlite-memory-mcp
claude mcp add --scope user sqlite_unified -- sqlite-memory-unified
Supported hosts
- Claude Desktop
- Claude Code
- Codex
Quick install
pip install sqlite-memory-mcp && claude mcp add --scope user sqlite_unified -- sqlite-memory-unifiedInformation
- Pricing
- free
- Published





