RAGSync MCP server adds semantic search over local docs and websites via YAML config
RAGSync is a configuration-driven MCP server that lets LLMs perform semantic search over local folders, PDFs, Markdown files, and web pages using a single YAML file and no code.
Score breakdown
RAGSync removes the need to write any ingestion or indexing code to give an LLM semantic search over a custom knowledge base, replacing what would otherwise be a bespoke RAG pipeline with a single YAML config.
- 01Indexes local folders (text, PDF, Markdown) and web pages via a single YAML config file — no code required.
- 02Five stable MCP tools: `search`, `list_sources`, `get_document`, `get_index_status`, and `reindex`.
- 03Live reload: filesystem watching and polling keep the index current; config changes apply without a restart.
RAGSync (`jsbroks/ragsync-mcp`) is a configuration-driven RAG MCP server designed to ingest, watch, and search arbitrary knowledge sources behind a stable tool surface. The entire setup is defined in a single YAML config file that specifies sources, chunking strategy, embedding model, and vector store — no application code is required. It supports local folders containing text, PDF, and Markdown files as well as web pages, and is not limited to a single file type or format.
The server includes live-reload capabilities: filesystem watching and polling keep the index current as source content changes, and edits to the config file itself are applied without restarting the server.
The server includes live-reload capabilities: filesystem watching and polling keep the index current as source content changes, and edits to the config file itself are applied without restarting the server. For embeddings, local `fastembed` works out of the box with no API key, with the option to swap in OpenAI or Voyage embeddings on a per-source basis. The MCP tool surface is intentionally stable — five source-agnostic tools (`search`, `list_sources`, `get_document`, `get_index_status`, `reindex`) that remain unchanged as new sources are added. Installation is available via `uvx` with no clone or install step, and the server is compatible with Cursor, Claude Desktop, Claude Code, Windsurf, and other MCP clients.
Key facts
- 01Indexes local folders (text, PDF, Markdown) and web pages via a single YAML config file — no code required.
- 02Five stable MCP tools: `search`, `list_sources`, `get_document`, `get_index_status`, and `reindex`.
- 03Live reload: filesystem watching and polling keep the index current; config changes apply without a restart.
- 04Local `fastembed` embeddings work out of the box with no API key; OpenAI or Voyage can be swapped in per source.
- 05Installable via `uvx` with no clone or install step.
- 06Compatible with Cursor, Claude Desktop, Claude Code, and Windsurf MCP clients.
- 07Config paths are resolved relative to the config file's directory, enabling repo-relative paths like `path: ./docs`.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 17, 2026 · 10:39 UTC. How this works →