MCP servers that search the web
Give your AI live web search results with sources.
10 servers · Last updated June 17, 2026
TL;DR: These servers return fresh search results so your agent can answer with current information and cite sources, instead of relying on stale training data. They differ on index (general vs developer-focused), whether they return full page content or just snippets, and pricing model.
Bottom line: if you only try one, Firecrawl is the most popular, verified option for this (6,562★). 9 more compared below.
Compare 10 servers
| Server | Transport | Auth | Verified | Stars | Tools for this |
|---|---|---|---|---|---|
| Firecrawl | Local (stdio) | API key | 6,562 | firecrawl_search | |
| Exa Search | Local (stdio) | API key | 4,573 | web_search_exa, web_search_advanced_exa, company_research_exa +1 | |
| Perplexity | Local (stdio) | API key | 2,299 | perplexity_search, perplexity_research | |
| Tavily | Local (stdio) | API key | 2,100 | tavily-search | |
| Apify Actors | Local (stdio) | API key | 1,336 | search-actors, search-apify-docs | |
| DuckDuckGo Search | Local (stdio) | No auth | 1,236 | search | |
| Brave Search | Local (stdio) | API key | 1,179 | brave_web_search, brave_local_search, brave_image_search +3 | |
| SearXNG | Local (stdio) | No auth | 896 | searxng_web_search, searxng_search_suggestions | |
| Jina AI Reader & Search | Remote (HTTP) | API key | 730 | search_web, search_images, search_arxiv +1 | |
| Kagi Search | Local (stdio) | API key | 417 | kagi_search_fetch |
The servers
Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.
Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.
Official Perplexity MCP server — web search, ask, deep research, and reasoning via Sonar models.
Production-ready MCP server for real-time web search, content extraction, site mapping, and crawling.
Official Apify MCP server — search and scrape the web via thousands of ready-made Actors; local or remote.
Popular no-API-key MCP server for DuckDuckGo web search plus page fetching and parsing.
Official Brave Search MCP server: web, local, image, video, news search and AI summarization.
Private metasearch via your own SearXNG instance — popular community MCP server for self-hosted search.
Official Jina AI remote MCP server — read web pages as Markdown and run grounded web search over HTTP.
Official Kagi MCP server (Python/uvx) — privacy-first web search and URL/video summarization.
Use these in a stack
FAQ
Which web-search MCP is best for AI agents?
For citations and clean snippets, API-first engines (Brave, Tavily, Exa) work well; Exa skews developer/semantic. Compare them head-to-head on the directory before committing.
Do these need an API key?
Most do — you'll provision a key from the search provider and drop it in the config. A few self-hosted options (SearXNG) avoid that.