MCP servers for browser and search
Last verified: 2026-09
Three different jobs get mixed up here. Search finds URLs (Brave, Exa, Tavily). Fetch downloads one public page as text. A browser (Playwright, Puppeteer, Browserbase) runs JavaScript and can click. Install one of each at most. Two search APIs just double the bill.
What to look for
- Public HTML → Fetch. Needs JS or a login → Playwright. Need to discover pages → one search API.
- Fetch can hit internal hostnames from your laptop (SSRF). Do not run it on a box that can see staging metadata.
- Playwright is the current browser default; Puppeteer is the older official example.
Side-by-side
| Server | Official? | Needs a secret? | Best for |
|---|---|---|---|
| Fetch | Yes | Yes | Fetch a URL and turn the page into readable text. |
| Brave Search | Yes | Yes | Web search via the Brave Search API. |
| Puppeteer | Yes | No | Drive a headless Chrome: click, type, screenshot. |
| Playwright | No | No | Microsoft's browser automation MCP for Chromium, Firefox, and WebKit. |
| Firecrawl | No | Yes | Crawl and scrape sites into clean markdown. |
| Browserbase | No | Yes | Cloud browsers you do not run locally. |
| Exa | No | Yes | Neural web search aimed at agents. |
| Tavily | No | Yes | Search API built for LLM retrieval. |
| Wikipedia | No | No | Look up Wikipedia pages without scraping. |
| EverArt | Yes | Yes | Generate images through EverArt. |
| arXiv | No | No | Search and read arXiv papers. |
All 11 servers
OfficialFetchFetch a URL and turn the page into readable text.
OfficialBrave SearchWeb search via the Brave Search API.
OfficialPuppeteerDrive a headless Chrome: click, type, screenshot.
Browser & searchPlaywrightMicrosoft's browser automation MCP for Chromium, Firefox, and WebKit.
Browser & searchFirecrawlCrawl and scrape sites into clean markdown.
Browser & searchBrowserbaseCloud browsers you do not run locally.
Browser & searchExaNeural web search aimed at agents.
Browser & searchTavilySearch API built for LLM retrieval.
Browser & searchWikipediaLook up Wikipedia pages without scraping.
OfficialEverArtGenerate images through EverArt.
Browser & searcharXivSearch and read arXiv papers.
FAQ
Why not give the model a full Chrome profile?
Then it can see every logged-in session. Use a clean browser and public sites first.
Is Firecrawl better than Fetch?
For multi-page crawls and clean markdown, yes. For one URL, Fetch is enough and has no API key.
Do I need Wikipedia MCP if I have Brave?
No. Brave can find the page; Fetch can read it.