Local RAG Search

Name: Local RAG Search
Availability: InStock
Author: nkapila6

Supports UI

by nkapila6

Privacy-first, local RAG-based web search across 9+ engines without requiring API keys.

0 stars

0 views

Works in:CursorGoose

Exposes:Tools

View on GitHub Docs

What it does

Provides a local RAG (Retrieval-Augmented Generation) system for web research. It performs multi-engine searches, uses Google's MediaPipe for local embeddings to rank results by semantic similarity, and returns extracted markdown content to the AI.

Tools

deep_research: Comprehensive research across multiple engines simultaneously.
deep_research_google: Focused deep dive using Google's index.
deep_research_ddgs: Privacy-focused research via DuckDuckGo.
rag_search_ddgs: Quick single search using DuckDuckGo.
rag_search_google: Quick single search using Google.

Installation

Using uvx (recommended):

{
  "mcpServers": {
    "mcp-local-rag": {
      "command": "uvx",
      "args": ["--python=3.10", "--from", "git+https://github.com/nkapila6/mcp-local-rag", "mcp-local-rag"]
    }
  }
}

Supported hosts

Confirmed for Claude Desktop, Cursor, and Goose.

Quick install

uvx --python=3.10 --from git+https://github.com/nkapila6/mcp-local-rag mcp-local-rag

Information

Pricing: free
Published: 4/17/2026
stars

Related Apps

Choose your AI client and follow the steps below.

Goose

Configure using uvx command.

Cursor

Add to mcp.json via uvx as described in README.

Claude Desktop

{"mcpServers": {"mcp-local-rag": {"command": "uvx", "args": ["--python=3.10", "--from", "git+https://github.com/nkapila6/mcp-local-rag", "mcp-local-rag"]}}}