searxng

name	searxng
version	1.0.0
description	Privacy-respecting metasearch engine aggregating results from 96+ search engines. Use the MCP server as the primary interface for web searches. Provides 2 tools via mcp__searxng__* prefixes including web search and URL content extraction.
tags	["search","searxng","web","privacy","metasearch","mcp"]
metadata	{"clawdbot":{"emoji":"🔍","requires":{"bins":[]}}}
env	{"SEARXNG_BASE_URL":{"description":"URL of the SearXNG server","required":true}}

SearXNG Web Search Skill

Privacy-respecting metasearch engine aggregating results from 96+ search engines. Use the MCP server as the primary interface — fall back to direct API for advanced queries.

MCP Server (Primary Interface)

The SearXNG MCP server is registered as searxng in ~/.mcp.json and provides 2 tools via mcp__searxng__* prefixes.

MCP Tools — Quick Reference

Tool	Purpose
mcp__searxng__searxng_web_search	Search the web (query, pageno, time_range, language, safesearch)
mcp__searxng__web_url_read	Fetch and extract content from a URL (with pagination/sections)

When to use MCP vs Direct API

Use MCP tools when...	Use direct API when...
General web searching	Searching specific categories (images, news, videos)
Reading URL content	Using specific engines (e.g., only github, arxiv)
Simple paginated results	RSS feed output
	CSV export
	Accessing suggestions, infoboxes, answers

MCP Server Location

• Binary: mcp-searxng (globally installed via npm, package mcp-searxng)
• Config: ~/.mcp.json → searxng entry
• Env vars: SEARXNG_URL

Configuration

Load environment from .openclaw/workspace/skills/searxng/.env:

• SEARXNG_BASE_URL — Server URL (default: http://localhost:8090)

Direct API (Fallback)

Search Endpoint

GET http://localhost:8090/search?q={query}&format=json

Query Parameters

Parameter	Type	Values	Description
q	string	any	Search query (required)
format	string	json, rss, csv, html	Response format
categories	string	comma-separated	Category filter
engines	string	comma-separated	Specific engines to query
language	string	ISO code (en, fr, all)	Result language
pageno	int	1+	Page number (default 1)
time_range	string	day, week, month, year	Recency filter
safesearch	int	0 (off), 1 (moderate), 2 (strict)	Content filtering

Example Queries

BASE="http://localhost:8090"

# Basic search
curl -s "$BASE/search?q=docker+tutorial&format=json" | python3 -m json.tool

# Search with time range (last week)
curl -s "$BASE/search?q=kubernetes+release&format=json&time_range=week"

# Search specific category
curl -s "$BASE/search?q=cat+photos&format=json&categories=images"

# Search specific engines only
curl -s "$BASE/search?q=python+asyncio&format=json&engines=stackoverflow,github"

# News search
curl -s "$BASE/search?q=AI+regulation&format=json&categories=news&time_range=month"

# Scientific papers
curl -s "$BASE/search?q=transformer+architecture&format=json&categories=science"

# Page 2 of results
curl -s "$BASE/search?q=rust+programming&format=json&pageno=2"

# RSS feed
curl -s "$BASE/search?q=open+source+CRM&format=rss"

JSON Response Structure

{
  "query": "search term",
  "number_of_results": 1234,
  "results": [
    {
      "url": "https://example.com/page",
      "title": "Result Title",
      "content": "Snippet description text...",
      "engine": "google",
      "engines": ["google", "duckduckgo", "bing"],
      "positions": [1, 3, 2],
      "score": 15.5,
      "category": "general",
      "publishedDate": "2026-03-27T12:21:00",
      "thumbnail": "",
      "img_src": ""
    }
  ],
  "answers": [],
  "corrections": [],
  "infoboxes": [],
  "suggestions": ["related term 1", "related term 2"],
  "unresponsive_engines": []
}

Result Fields by Category

General results: url, title, content, engines, score, publishedDate

News results add: source, metadata (e.g., "9 days ago | Reuters")

Image results add: thumbnail_src, img_src, resolution, img_format, filesize, source

Video results add: thumbnail, length (duration), author

Available Categories

Category	Description
general	Web search (default)
images	Image search
videos	Video search
news	News articles
music	Music and audio
files	File/torrent search
it	IT/programming
science	Scientific papers
social media	Social platforms
map	Map/location search
q&a	Q&A sites
repos	Code repositories
packages	Software packages
wikimedia	Wikimedia content

Key Engines

General/Web

google, bing, duckduckgo, brave, startpage, wikipedia, wikidata

IT / Development

github, docker hub, pypi, mdn, arch linux wiki, stackoverflow, askubuntu, superuser

Science

arxiv, google scholar, pubmed, semantic scholar

News

google news, bing news, reuters, yahoo news, wikinews

Images

google images, bing images, flickr, pexels, pinterest, unsplash, deviantart

Videos

youtube, google videos, dailymotion, vimeo

Social

lemmy, mastodon, tootfinder

Instance Config

Get full engine/category list:

curl -s http://localhost:8090/config | python3 -m json.tool

Rate Limiting Notes

• SearXNG itself has no API rate limiting (limiter is disabled on this instance)
• Upstream engines (Google, Bing, etc.) may throttle if queried too aggressively
• Best practice: add client-side delays for bulk queries, use pagination instead of re-searching
• Cache results when possible (SearXNG caches internally via Valkey)

Web UI

Access the visual search interface at: http://localhost:8090

MCP Update

npm install -g mcp-searxng
# Then restart Claude Code