michielinksee/linksee-memory

What it does

linksee-memory is an MCP (Model Context Protocol) server that provides persistent, cross-agent memory management for AI coding assistants. It stores and retrieves information using a 6-layer structure with Ebbinghaus-style forgetting curves to simulate natural memory decay. The tool includes drift detection to alert users when code decisions diverge from previously documented patterns, ensuring consistency across projects.

How to set up

Installation requires setting up the MCP server from the GitHub repository. The tool uses a single SQLite database file that can be shared across multiple AI platforms including Claude, Cursor, Codex, and Gemini. Configuration involves pointing your AI tools to the local MCP server endpoint. Detailed setup instructions are available in the repository documentation.

Key features

• 6-layer memory structure for organized information storage
• Ebbinghaus forgetting curve implementation for realistic memory decay
• Drift detection to catch code divergence from past decisions
• Single SQLite file for seamless multi-platform sharing
• Local-first architecture with no external account requirements
• Support for Claude, Cursor, Codex, and Gemini