mcp-local-rag
mcp-local-rag is hosted online, so all tools can be tested directly either in theTools tabor in theOnline Client.
If you are the rightful owner of mcp-local-rag and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcpreview.com.
mcp-local-rag is a primitive RAG-like web search model context protocol server that runs locally without the need for APIs.
mcp-local-rag is a local server designed to enhance language models by providing real-time web search capabilities. It operates without relying on external APIs, making it a self-contained solution for fetching and processing web data. The server uses DuckDuckGo for search queries and Google's MediaPipe Text Embedder for generating embeddings. It ranks search results based on similarity to the query and extracts relevant context from the top results. This context is then returned to the language model in markdown format, allowing for more informed and up-to-date responses. The server can be run using Python with uv, cloned and run locally, or deployed using Docker. It is compatible with various MCP clients and has been tested with platforms like Claude Desktop, Cursor, and Goose.
Features
- Local Operation: Runs entirely on local machines without the need for external APIs.
- Real-time Web Search: Utilizes DuckDuckGo to fetch live search results.
- Context Extraction: Extracts and returns relevant context from web pages in markdown format.
- Embeddings and Ranking: Uses Google's MediaPipe Text Embedder to rank search results based on query similarity.
- Flexible Deployment: Can be run using Python, Docker, or cloned locally.
Tools
rag_search
Search the web for a given query. Give back context to the LLM with a RAG-like similarity sort. Args: query (str): The query to search for. num_results (int): Number of results to return. top_k (int): Use top "k" results for content. Returns: Dict of strings containing best search based on input query. Formatted in markdown.