Deep Search Semantic Memory
Enables local semantic search and long-term memory retrieval for large-scale codebases and documentation within Claude.
Deep Search bridges the gap between Claude's working memory and massive project repositories by integrating with the qmd semantic search engine. It automatically detects when a project exceeds standard context limits and offers a non-intrusive way to index documentation and source code locally. By using BM25 and vector embeddings, it allows Claude to pinpoint specific architectural decisions, historical PRDs, or relevant code fragments across thousands of files, ensuring high-quality responses even in enterprise-scale environments.
Key Features
01Automated indexing sync to prevent stale search results
02Local semantic search using BM25, Vector embeddings, and LLM re-ranking
03Multi-tier memory system bridging working memory and long-term storage
04Zero-config MCP server integration for seamless Claude communication
05Automatic detection of oversized projects and documentation folders
0645 GitHub stars
Use Cases
01Navigating legacy codebases with hundreds of source files and architectural specs
02Synchronizing newly generated documentation into a searchable local knowledge base
03Retrieving specific historical design decisions from large documentation folders