MCP RAG Starter Kit

Example Use Cases

• 1

  Search and retrieve relevant passages from a local document collection using vector search.

• 2

  Ready-made prompt templates to perform summarization, keyword extraction, translation, and question-answering over retrieved documents.

• 3

  Expose a local RAG tool to Claude Desktop or other MCP-capable clients for interactive use.

Description

This repository implements a starter RAG server using the MCP (Model Context Protocol) to index and search documents and to run text-processing prompt templates (summary, keyword extraction, translation, Q&A). It supports multilingual embeddings including Korean and integrates with OpenAI and Hugging Face models. Documents placed in the docs directory are automatically indexed into a vector store and reindexed on changes. It also includes configuration guidance to integrate the server as a tool in Claude Desktop.