SearXNG Websearch MCP

This MCP server enables your local LM Studio setup to perform private web searches through a local SearXNG instance. It supports concurrent searches, result caching, and advanced filtering, all while keeping queries on your machine for privacy and speed.

How to use

You will use this MCP server from LM Studio to perform web searches through your local SearXNG instance. Start the SearXNG service, then configure LM Studio to load the MCP server so your LLMs can issue search requests. You can run single searches or multiple searches in parallel, with results cached to improve performance over time.

How to install

# Prerequisites
- Node.js 18 or higher
- Docker and Docker Compose
- LM Studio

# 1. Install and Build
git clone https://github.com/Pascalrjt/SearXNG-Websearch-MCP.git
cd SearXNG-Websearch-MCP

npm install
npm run build

# 2. Start SearXNG locally (Docker)
npm run docker:up

# Verify SearXNG is running in a browser
# Open http://localhost:8080

# 3. Configure LM Studio to load the MCP server (see next section)

Configuration and usage in LM Studio

Add the following MCP server entry to your LM Studio configuration. This defines a local stdio MCP that runs the built MCP server script.

{
  "mcpServers": {
    "websearch-searxng": {
      "command": "node",
      "args": ["/path/to/SearXNG-Websearch-MCP/dist/index.js"],
      "env": {}
    }
  }
}

Start and verify the MCP server in LM Studio

Restart LM Studio after you add the MCP server configuration so LM Studio detects the new server.

Then verify that the MCP server is reachable by triggering a search from your LM Studio LLM and watching the results.

Usage examples in LM Studio

Basic search: Ask your LLM to perform a web search for a topic, then summarize the findings. The MCP server will route the query to your local SearXNG instance and return results.

Troubleshooting and tips

If the MCP server does not connect in LM Studio, ensure the build completed and the dist/index.js file exists, and confirm the path in the LM Studio config is absolute. Restart LM Studio if needed.

Security and privacy

All searches run through your local SearXNG instance. No data leaves your machine, and you retain full control over privacy settings and search engines used by SearXNG.

Performance considerations

Use the automatic caching to reduce repeated API calls. Enable concurrent searches via multi_search to gather multiple results efficiently. Apply domain filtering and maxResults to keep responses focused and fast.

System prompt and optimization

A comprehensive system prompt is included to guide your LLM in using the web search tools effectively, including date awareness, cross-source verification, and caching management.

Available tools

web_search

Performs a single web search with advanced filtering options such as language, time range, categories, domain filtering, and deduplication.

multi_search

Executes multiple searches concurrently to speed up information gathering and comparison.

clear_cache

Clears the internal cache of search results to force fresh results.