Web research MCP servers for deep, cited answers
MCP servers for web research: search APIs, page fetchers, and deep research agents that gather sources and return cited answers to your AI client.
What to know about Web research MCP servers
Web research servers go beyond a single search query. They combine a search API with page fetching, and in some cases run a multi-step loop that searches, reads, refines the question, and searches again before returning a summary with citations. Providers such as Exa, Tavily, Perplexity, Brave, and Kagi each have servers, alongside self-hosted options built on SearXNG.
The distinction that matters is who does the reasoning. Plain search servers return links and snippets, and your client's model decides what to read next. Deep research servers do that loop on the provider's side and hand back a finished report, which saves your context but gives you less control over sources.
For trustworthy results, prefer servers that return URLs with every claim and that can fetch the full page when snippets are not enough. Then spot-check a few sources, especially for anything recent or contested.
Web research MCP servers
The 12 strongest of 4173 matching servers. Servers that name the topic directly come first, then by trust score.
- 75Trust score 75 of 100, Good
SearXNG Search
by ihor-sokoliuk · TypeScript
MCP server for SearXNG — privacy-respecting web search with pagination, URL reading
RegistryLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 25 days ago.
72Trust score 72 of 100, GoodFree Search
by sweetcornna · Python
No-API-key web search, page fetching and document reading across many engines. Local-first.
RegistryLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 16 days ago.- 70Trust score 70 of 100, Good
You.com Web Access & AI
by youdotcom-oss · TypeScript
Web search, AI agent, and content extraction via You.com APIs
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 14 days ago. - 69Trust score 69 of 100, Good
Official SerpApi MCP server for Google, Bing, and other search engines.
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 23 days ago. - 69Trust score 69 of 100, Good
Open-source MCP server for AI agents: web search, content extraction, and library docs.
RegistryLicenseQualityMaintenanceLicense: A. Apache-2.0 is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 11 days ago. - 68Trust score 68 of 100, Good
web-search
by keenableai · Shell
Live web search and clean-markdown page fetch over the Keenable web index.
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 25 days ago. - 68Trust score 68 of 100, Good
Jina Reader
by pipeworx-io · JavaScript
Jina AI Reader/Search MCP — turn any URL into clean LLM-ready markdown, plus web search.
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 9 days ago. - 68Trust score 68 of 100, Good
Serpapi
by pipeworx-io · JavaScript
SerpApi MCP — wraps SerpApi (serpapi.com) search engines
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 9 days ago. - 67Trust score 67 of 100, Good
Brave Search
by pipeworx-io · JavaScript
Brave Search MCP — independent web index (no Google/Bing dependency)
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 9 days ago. - 67Trust score 67 of 100, Good
Exa
by pipeworx-io · JavaScript
Exa MCP — neural/semantic web search + content retrieval (exa.ai)
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 9 days ago. - 67Trust score 67 of 100, Good
Marginalia
by pipeworx-io · JavaScript
Marginalia MCP — independent web search over a non-commercial, non-SEO index.
RegistryRemoteLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 9 days ago.
66Trust score 66 of 100, Goodsearch-rotation
by localfoundry · TypeScript
MCP server for web search and page extraction across the free tiers of several providers
RegistryLicenseQualityMaintenanceLicense: A. MIT is a permissive open source license.Quality: Not graded. No README or tool list has been captured yet.Maintenance: A. Last commit or release 6 days ago.
How to install one
- Compare the shortlist. SearXNG Search leads this topic. Open two or three listings and compare their tools, transport, and required API keys.
- Copy the config for your client. Each listing has an install guide, for example Claude Desktop, Claude Code, Cursor, or VS Code. The client guides show where each config file lives.
- Scope credentials, then restart. Use the narrowest token or folder the server needs, restart the client, and check that the new tools appear.
Web research MCP questions
What is the difference between a search server and a research server?
A search server returns results for one query. A research server runs several searches and reads pages on its own, then returns a synthesised answer. Research servers are slower and usually cost more per call.
Are there free web search MCP servers?
SearXNG and DuckDuckGo servers need no API key. Brave, Tavily, and Exa offer free tiers that cover personal use. See the web search use case for the strongest options.
How do I make sure the answers are cited?
Pick a server that returns source URLs, and ask the model to cite a link for each claim. If a claim has no link, treat it as the model's own inference.