Best MCP servers for web scraping
20 working options, checked Nov 8, 2026.
Short answer: the top picks for web scraping right now are Firecrawl, XCrawl MCP and browser-scraper. We list 20 options here: 13 work without sign-in, 2 need a sign-in and 5 run on your own machine.
A web scraping MCP server lets an AI assistant open a live page and read it as clean text, instead of guessing from what it remembers. Most return Markdown or JSON, and the stronger ones handle JavaScript-heavy sites and whole-site crawls.
Top picks
-
Firecrawl
Industrial-strength scraping: crawl entire sites into clean markdown.
-
XCrawl MCP
Let an agent scrape, crawl, map and search the live web and get clean Markdown or JSON back.
-
browser-scraper
Local MCP server that gives Claude, Cursor or Codex a stealth Chrome over raw CDP: Markdown, metadata and clickable snapshot refs, closed shadow DOM included.
-
Irmu
Web data infrastructure for AI agents: scraping, screenshots, AI answers and travel data through one hosted MCP server.
-
FastAF
Structured web data for AI agents: Amazon, Google search and more as JSON.
Tools include
scrape_url,google_search,google_maps_places,amazon_product,search_news
All 20 compared
| Server | What it does | Setup | Tools |
|---|---|---|---|
| Firecrawl | Industrial-strength scraping: crawl entire sites into clean markdown. | Runs locally | – |
| XCrawl MCP | Let an agent scrape, crawl, map and search the live web and get clean Markdown or JSON back. | Runs locally | – |
| browser-scraper | Local MCP server that gives Claude, Cursor or Codex a stealth Chrome over raw CDP: Markdown, metadata and clickable snapshot refs, closed shadow DOM included. | Runs locally | – |
| Irmu | Web data infrastructure for AI agents: scraping, screenshots, AI answers and travel data through one hosted MCP server. | Hosted, sign-in | – |
| FastAF | Structured web data for AI agents: Amazon, Google search and more as JSON. | Hosted, no sign-in | 23 |
| GetMyData | Ask your AI for a spreadsheet of public web data. | Hosted, no sign-in | 16 |
| ScrapeMole Reader | Pay-per-call web reading for agents: any page as clean Markdown, link-preview metadata, RSS/Atom feeds (x402, USDC on Base). | Hosted, no sign-in | 5 |
| AgentSearch MCP | Web search, page extraction and rendering for AI agents through one MCP server. | Runs locally | – |
| WebExtract | Any public web page to clean Markdown or fixed-schema JSON for agents. Pay per call, no signup. | Hosted, no sign-in | 2 |
| NanoParse | Clean Markdown from any URL, with Litmus source-intelligence signals. | Hosted, no sign-in | 2 |
| WebRead | Read any public web page as clean Markdown for LLMs. Respects robots.txt. | Hosted, no sign-in | 1 |
| Avo Cado Read | Turn any public web page into clean Markdown for AI agents, pay-per-call via x402. | Hosted, no sign-in | 2 |
| ScrapeAgent | Pay-per-use web scraping for AI agents — $0.01 USDC, no API key needed | Hosted, no sign-in | 1 |
| switch402 | Web scraping, search, screenshots and PDF to text for agents, paid per call over x402. No signup. | Hosted, no sign-in | 9 |
| ReefAPI MCP | One MCP server for 250+ live web-data APIs | Hosted, no sign-in | 5 |
| PageWatch | Extract prices, lists, tables and events from static web pages and watch them for changes. | Hosted, no sign-in | 3 |
| Product Page Extractor | Turn a product page URL, or raw HTML, into clean JSON: price, currency, stock, SKU/GTIN/MPN, images, rating and every offer, with price_source attribution. | Hosted, no sign-in | 1 |
| Agent Extraction API (FreeTier Agent API) | Pay-per-call API and MCP server for AI agents: invoice PDF to JSON, clean web pages, contact extraction. | Hosted, no sign-in | 74 |
| Fetch | Fetch web pages and convert them to clean, readable text for your agent. | Runs locally | – |
| ScrapeAtlas | Unified scraper API for every social platform — public profiles, posts, videos and conversations via one social media data API. | Hosted, sign-in | – |
How to choose
- Single pages or whole sites? Reading one URL is cheap and simple. Crawling a site needs a server built for it, with limits you can set.
- Does the site need a real browser? Plain fetchers miss content that loads with JavaScript. Pick one that renders pages if your targets are modern web apps.
- How do you want to pay? Some use an API key and a plan, others charge per call with no account, and local options cost only your own machine time.
How to connect one
Here is Irmu as an example. In Claude Code, run:
claude mcp add --transport http irmu https://app.irmu.com/mcpFor other apps, follow our guides for Claude, ChatGPT, Cursor and VS Code.
How we picked these
We read each listing's tools and put the closest fit for this job first. Hosted servers must pass a real MCP handshake in our daily check to stay on this page, and local servers are reviewed from their source or package. That shows a server responds, not how well it works for you, and we don't run every tool in production.
Frequently asked questions
Which of these MCP servers work without signing in?
FastAF, GetMyData, ScrapeMole Reader, WebExtract and NanoParse are among the 13 that answered our latest live check with no account or key.
Can an MCP server scrape any website?
No. Public pages usually work, but login walls, bot protection and robots.txt rules limit what any scraper can reach. Check a site's terms before scraping it at volume.
What is the difference between fetching and scraping in MCP?
A fetch tool reads one URL and returns its text. A scraping or crawling tool goes further: it can render JavaScript, follow links across a site and return structured data.
How often is this list updated?
Every hosted server here is re-checked daily with a real MCP handshake, and a server drops off this page while it fails that check. Last check: Nov 8, 2026.