MCP Now
mcp-local-rag

mcp-local-rag

by nkapila6
GitHub

"primitive" RAG-like web search model context protocol (MCP) server that runs locally using Google's MediaPipe Text Embedder and DuckDuckGo Search. ✨ no APIs required ✨.

http
model
clone
config
path

mcp-local-rag

"primitive" RAG-like web search model context protocol (MCP) server that runs locally. ✨ no APIs ✨

Installation instructions

  1. You would need to install uv: https://docs.astral.sh/uv/

If you do not want to clone in Step 2.

Just paste this directly into Claude config. You can find the configuration paths here: https://modelcontextprotocol.io/quickstart/user

1{
2    "mcpServers": {
3        "mcp-local-rag":{
4            "command": "uvx",
5            "args": [
6            "--python=3.10",
7            "--from",
8            "git+https://github.com/nkapila6/mcp-local-rag",
9            "mcp-local-rag"
10            ]
11        }
12    }
13}

Otherwise:

  1. Clone this GitHub repository (OPTIONAL, can be skipped with above config)
1git clone https://github.com/nkapila6/mcp-local-rag
  1. Add the following to your Claude config. You can find the configuration paths here: https://modelcontextprotocol.io/quickstart/user
1{
2  "mcpServers": {
3    "mcp-local-rag": {
4      "command": "uv",
5      "args": [
6        "--directory",
7        "<path where this folder is located>/mcp-local-rag/",
8        "run",
9        "src/mcp_local_rag/main.py"
10      ]
11    }
12  }
13}

Example use

On prompt

When asked to fetch/lookup/search the web, the model prompts you to use MCP server for the chat.

In the example, have asked it about Google's latest Gemma models released yesterday. This is new info that Claude is not aware about.

Result

The result from the local rag_search helps the model answer with new info.