Here's a summary of the README content, restructured for clarity and within the specified word count:

mcp-local-rag

This project provides a local, "primitive" Retrieval-Augmented Generation (RAG)-like web search model using the Model Context Protocol (MCP). It operates locally without requiring external APIs.

Installation

1. Install uv from: https://docs.astral.sh/uv/

Configuration Options:

• Direct Configuration: Paste the provided JSON directly into your Claude configuration (found at https://modelcontextprotocol.io/quickstart/user). This method avoids cloning the repository. The JSON configures the mcp-local-rag server to use uvx to install and run the project directly from the GitHub repository.
• Cloning and Configuration: Alternatively, clone the repository using git clone https://github.com/nkapila6/mcp-local-rag. Then, add the second provided JSON configuration to your Claude settings, adjusting the --directory path to point to the cloned repository location. This config uses uv to run the src/mcp_local_rag/main.py script.

Example Use

When prompted to search the web, the model will request the use of the MCP server. The local rag_search function then provides the model with relevant information to answer the query.