agents-libagents-lib
AgentsMCP ServersWorkflowsTutorialsComparisonsTemplatesNewsAbout
🔌

SearXNG MCP

Search•⭐ 380

Privacy-focused meta-search engine with support for multiple search engines.

GitHubWebsite →
Claude DesktopCursor

Overview

Privacy-focused meta-search engine with support for multiple search engines.

Setup

Run with npx:

npx -y @modelcontextprotocol/server-searxng

Configuration

SEARXNG_BASE_URL environment variable

Documentation

SearXNG MCP

Overview

SearXNG MCP is a Model Context Protocol server that provides access to SearXNG, a privacy-focused meta-search engine. SearXNG aggregates results from multiple search engines (Google, Bing, DuckDuckGo, Wikipedia, and many more) while protecting user privacy by not tracking searches or building profiles.

The MCP integration allows AI agents to perform web searches through SearXNG instances, benefiting from aggregated results across multiple engines while maintaining privacy. You can use public instances or self-host your own for full control.

Features

  • Meta-Search: Aggregates results from 50+ search engines
  • Privacy-Focused: No tracking, no cookies, no profiling
  • Self-Hostable: Run your own instance for full control
  • Customizable: Choose which engines to use
  • Category Support: Search web, images, videos, news, science, and more
  • Safe Search: Configurable content filtering
  • Language Support: Multi-language search results

Installation

npx -y @modelcontextprotocol/server-searxng

Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "searxng": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-searxng"],
      "env": {
        "SEARXNG_BASE_URL": "https://searx.example.com"
      }
    }
  }
}

Using Public Instances

You can use public SearXNG instances. Find a list at searx.space:

{
  "mcpServers": {
    "searxng": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-searxng"],
      "env": {
        "SEARXNG_BASE_URL": "https://searx.be"
      }
    }
  }
}

Self-Hosting

docker run -d --name searxng \
  -p 8080:8080 \
  -v /path/to/searxng:/etc/searxng \
  searxng/searxng

Available Tools

ToolDescription
searchPerform a web search
search_imagesSearch for images
search_videosSearch for videos
search_newsSearch for news articles
search_scientificSearch scientific papers

Usage Examples

Basic Web Search

results = search(query="AI agent frameworks comparison 2026")
for result in results:
    print(f"{result.title}: {result.url}")
    print(f"  {result.content}")

Category-Specific Search

# Search for images
images = search_images(query="machine learning architecture diagrams")

# Search for news
news = search_news(query="latest AI breakthroughs")

# Search scientific papers
papers = search_scientific(query="transformer models attention mechanism")

Filtered Search

results = search(
    query="python async programming",
    categories=["web", "programming"],
    language="en",
    safesearch=1,  # 0=off, 1=moderate, 2=strict
)

Pros

  • ✅ Privacy-First: No tracking or profiling
  • ✅ Aggregated Results: Multiple engines in one query
  • ✅ Self-Hostable: Full control over your instance
  • ✅ No API Key Required: Use public instances freely
  • ✅ Customizable: Choose which engines to use
  • ✅ Multi-Category: Web, images, videos, news, science

Cons

  • ❌ Public Instance Reliability: Public instances may be slow or unavailable
  • ❌ Rate Limiting: Some instances have rate limits
  • ❌ Less Comprehensive: May miss results from individual engines
  • ❌ Self-Hosting Overhead: Requires maintenance if self-hosted

When to Use

Use SearXNG MCP when:

  • You value privacy and don't want search tracking
  • You want aggregated results from multiple engines
  • You want to avoid API keys and costs
  • You're building privacy-conscious applications

Avoid SearXNG MCP when:

  • You need guaranteed uptime (self-host instead)
  • You need the most comprehensive results (use individual engines)
  • You need enterprise-grade support

Resources