# rag MCP server

> Enables AI clients to search and manage a local Markdown knowledge base using hybrid full-text and vector search.

- Repository: https://github.com/core-euler/rag-mcp
- Page: https://hostedmcp.org/servers/core-euler/rag-mcp
- Transport: —
- Stars: 2
- License: MIT

## FAQ

### How does the search work?

It uses a hybrid approach combining SQLite FTS5 for keyword matching and cosine similarity over vector embeddings, merged via Reciprocal Rank Fusion.

### Do I need a vector database?

No, the server uses a SQLite database for the full-text index and performs vector similarity via in-memory matrix multiplication using numpy.

### Can I use it with Obsidian?

Yes, the system is designed so that Markdown files remain ordinary files that you can edit in Obsidian or any other editor.

### What happens if I delete the index file?

The index can be rebuilt from your Markdown files.