Codemem

About

Codemem functions as a standalone Rust memory engine specifically crafted for AI coding assistants, operating as a single binary with zero runtime dependencies. It meticulously captures and stores everything an AI assistant discovers, from files read and symbols searched to edits made, ensuring that code repositories do not need repeated re-exploration across multiple sessions. This persistent memory allows AI assistants to build upon prior knowledge, enhancing their efficiency and intelligence, and offering sophisticated tools for code understanding and interaction.

Key Features

• Code-aware indexing using tree-sitter for structural extraction across 13 programming languages.
• 38 MCP tools for memory management, graph traversal, code search, and impact analysis via JSON-RPC.
• 9-component hybrid scoring for highly precise context retrieval and ranking.
• Graph-vector hybrid architecture combining HNSW vector search and petgraph knowledge graphs.
• Neuroscience-inspired memory consolidation cycles and self-editing capabilities with full provenance tracking.
• 1 GitHub stars

Use Cases

• Persisting AI assistant discoveries and conversational context across multiple coding sessions.
• Indexing and analyzing codebase architecture to build structural knowledge graphs for AI agents.
• Providing advanced memory management and context recall for AI coding assistants through a comprehensive toolset.