1
Exa's hosted neural search server with web search, page fetch and deep research
★ 5.0k · MIT · updated 2026-09-23
8.3/10
What it is The Exa MCP Server is Exa's official server for its AI-native search engine. It is hosted at mcp.exa.ai/mcp and open source on GitHub. It gives agents web search with clean page content, full-page fetching as Markdown and, optionally, advanced search and an agent tool for multi-step research. Key features - web_search_exa returns results with clean content -… Read more →
Pros
- Works without a key to start
- Hosted, many supported clients
- Full-page fetch as Markdown
- Active official repo
Cons
- Anonymous use is rate-limited
- Deep search and agent runs cost more
- Pay-as-you-go only
Pricing: Freemium · web-search research neural-search official hosted
2
Official Perplexity API server for cited web search, answers and deep research
★ 2.5k · MIT · updated 2026-08-27
7.9/10
What it is The Perplexity MCP Server is Perplexity's official server for its API platform. It gives an assistant Perplexity's real-time web search and answer engines, from quick cited answers to longer research runs. It runs locally from npm or through Perplexity's hosted endpoint at api.perplexity.ai/mcp. Key features - perplexity_search: ranked web results via the Search API - perplexity_ask: conversational… Read more →
Pros
- Answers with citations
- Search, ask, research and reason tools
- Official, MIT-licensed
- Hosted or local
Cons
- No free API allowance stated
- API in transition from Sonar to Agent API
- Research runs are slow and cost more
Pricing: Freemium · web-search research citations official answer-engine
3
Tavily's search, extract, map and crawl tools for agents, hosted or local
★ 2.4k · MIT · updated 2026-09-16
7.9/10
What it is The Tavily MCP Server is Tavily's official server for its search API built for AI agents. It gives assistants real-time web search, content extraction, site mapping and crawling. You can use Tavily's hosted endpoint or run it locally with npx. Key features - search: real-time web search tuned for LLMs - extract: pull clean content from given… Read more →
Pros
- 1,000 free credits a month
- Search and extraction together
- Hosted or local, several auth options
- Official, MIT-licensed
Cons
- Key in the URL can leak into logs
- Overlaps heavily with Firecrawl and Exa
- Crawl jobs use many credits
Pricing: Freemium · web-search extraction research official open-source
4
Official Brave server for web, news, image, video and local search plus summaries
★ 1.5k · MIT · updated 2026-09-24
7.8/10
What it is The Brave Search MCP Server is Brave's official open-source server for the Brave Search API, which runs on Brave's own independent web index. It gives agents web, news, image, video and local search, plus AI summaries and LLM-ready context. Key features - Web search with filters - News, image and video search - Local business and point-of-interest… Read more →
Pros
- Independent index
- News, image, video and local search
- Official and MIT-licensed
- Monthly free credits
Cons
- Snippets, not full page content
- Paid per request beyond free credits
- Requires Node.js 22+ locally
Pricing: Freemium · web-search news research official open-source
5
Community server that fetches YouTube captions, with optional TwelveLabs video analysis
★ 597 · MIT · updated 2026-07-21
7.0/10
What it is YouTube Transcript MCP is a community Model Context Protocol server by kimtaeyoon83 that pulls captions from YouTube videos so an assistant can summarise, quote or repurpose them. It is a small TypeScript project that runs locally with Node.js. Key features - get_transcript with language selection - Optional timestamps in the output - Filters sponsor segments and ads… Read more →
Pros
- Free, no API key for transcripts
- Filters sponsor segments
- Supports Shorts
- Simple setup
Cons
- Depends on YouTube caption access that can change
- Only videos with captions
- Community-maintained, no SLA
Pricing: Open source · youtube transcripts repurposing research community