jagoff/memo

What it does

jagoff/memo is a semantic memory server that runs locally on your machine, providing intelligent storage and retrieval of information without relying on external services. It uses hybrid search combining traditional and semantic methods, supports plain Markdown files for easy editing and version control, and can generate embeddings using either MLX (on Apple Silicon) or CPU-based processing.

The tool excels at detecting contradictions in stored information, maintaining complete edit history with time-travel capabilities, and synthesizing new insights from existing memories. It synchronizes across machines using Git, ensuring your knowledge base stays consistent and backed up.

How to set up

Installation requires Python and Git. Clone the repository from GitHub, install dependencies using pip, configure your embedding backend preference (MLX or CPU), and initialize your local Markdown storage directory. The MCP server can then be integrated with compatible AI tools and applications.

Key features

• Hybrid semantic and keyword search
• Plain Markdown storage for human-readability and version control
• MLX or CPU-based embeddings (no cloud dependency)
• Automatic contradiction detection
• Complete time-travel history of all changes
• Information synthesis and insight generation
• Cross-machine synchronization via Git
• Local-first privacy and control