mind-mem

What it does

mind-mem is a memory management system designed specifically for coding agents and AI assistants. It provides persistent, auditable, and contradiction-safe memory storage using a hybrid search approach combining BM25 full-text search and vector embeddings with Reciprocal Rank Fusion (RRF) ranking. The system supports multiple transport methods including HTTP, inbox, and daemon modes, making it flexible for various agent architectures.

How to set up

Install via pip as a Python package. Configure transport protocol (HTTP/inbox/daemon). Initialize SQLite database for memory storage. Register with your AI client (supports Claude, ChatGPT, and 15+ others via MCP protocol). Set up hash-of-code pipeline for automatic invalidation when dependencies change.

Key features

• Hybrid BM25+vector+RRF search for optimal retrieval accuracy
• Native MIC/MAP wire format support
• Contradiction detection and resolution
• Hash-of-code pipeline invalidation for dependency tracking
• Persona projection capabilities
• Dependency-ordered walkthrough
• 81 integrated MCP tools
• Compatible with 17+ AI clients including Claude Desktop, Cursor, and AI Studio
• Zero runtime dependencies
• SQLite-based persistent storage