This gives Claude local file search capabilities using a hybrid approach that combines BM25 keyword matching with semantic embeddings, all running offline on your machine. You'd reach for this when you want Claude to find relevant documents across your filesystem without sending anything to external services. The semantic layer means it can surface contextually related files even when exact keywords don't match, while BM25 handles precise term searches. Useful if you're working with large document collections and need Claude to pull context from your local knowledge base during conversations. Runs over stdio, so it integrates as a standard MCP server.