01Semantic code indexing via Tree-sitter AST parsing for accurate context
02Persistent cross-session memory for recalling AI decisions and code areas
03Live dashboard and dollar estimates for tracking token savings
043 GitHub stars
05Hybrid vector and BM25 search with code graph expansion and confidence scoring
06Achieves up to 94% token savings for AI coding agent interactions (benchmarked)
07Local-first MCP server with zero-cloud, zero-config setup and editor auto-detection