AI and agentsCapability overview

MCP Servers for Web Scraping: What Each Provider Offers

Bright Data, Oxylabs and ZenRows all publish official MCP servers that let Claude, Cursor and other agents fetch live web data. Here is what each one offers according to its documentation, how to connect it, and which fits which job.

Checked against provider docs 2 min read
how an MCP scraping call works
  1. 1You ask the agent — "Get the price of this product."
  2. 2The agent calls a tool on the provider's MCP server, with the URL.
  3. 3The provider fetches the page through its own network and unblocking.
  4. 4Data comes back as Markdown, HTML or JSON, and the agent answers.
How this page was made. Based on each provider's public documentation, checked on . We haven't run these through our hands-on tests yet. Our verdicts on each provider's core products are in the reviews.

At a glance

Capabilities side by side

Every cell links to the provider doc it came from. A dash means the docs don't mention it.

CapabilityBright DataOxylabsZenRows
Official MCP server✓ open source (MIT)✓ open source (MIT)✓ open source
Hosted (remote) server✓ mcp.brightdata.com✓ mcp.oxylabs.io✓ mcp.zenrows.com
Fetch a page as Markdown✓ scrape_as_markdown✓ Markdown, HTML or links✓ Markdown (default), text, HTML
JavaScript rendering—✓ full, HTML-only or none✓ js_render
Anti-bot unblocking✓ unlocker technology✓ CAPTCHA and geo-restrictions✓ on every tool
Search engine results✓ Google, Bing, Yandex✓ Google; AI web search—
Structured data✓ pre-built scrapers (Amazon, Booking.com, more)✓ Google and Amazon parsers; AI extraction✓ extract (JSON fields)
Browser actions (click, type, scroll)✓ scraping_browser_* tools✓ ai_browser_agent (prompt-driven)✓ browser_* tools
Crawl a whole site—✓ ai_crawler✓ crawl_* tools
Geo-targeting✓ bypasses geo-restrictions✓ 195+ countries✓ proxy_country
Works in Claude Desktop and Cursor✓✓✓
Free tier or trial5,000 requests/month1-week trial; 1,000 AI Studio credits5,000 credits, first billing period

Provider by provider

What each MCP server is built for

Bright DataBright Data MCP

Lets LLMs and agents search, scrape and extract web data in real time, with pre-built scrapers for sites such as Amazon and Booking.com and tools to drive a remote browser. Source

Tools it exposes · source
search_enginescrape_as_markdownscrape_as_htmlrun_scraperscraping_browser_navigatescraping_browser_click_refsession_stats
Free tier or trial
5,000 requests a month, shared with other Bright Data products
Billing
Pay-as-you-go after the free allowance
Our provider rating
4.7/5
Try Bright Data

Sponsored link. Ratings are independent.

Official MCP docs →
OxylabsOxylabs MCP (Web Scraper API and AI Studio tools)

Connects AI models to the web through Web Scraper API tools for any URL, Google Search and Amazon, plus AI Studio tools that scrape, crawl, search and control a browser from a prompt. Source

Tools it exposes · source
universal_scrapergoogle_search_scraperamazon_search_scraperamazon_product_scraperai_scraperai_crawlerai_browser_agentai_searchai_mapgenerate_schema
Free tier or trial
1-week Web Scraper API trial; 1,000 free AI Studio credits
Our provider rating
4.5/5
Try Oxylabs

Sponsored link. Ratings are independent.

Official MCP docs →
ZenRowsZenRows MCP

Gives MCP clients live web pages through scraping and cloud-browser tools, with anti-bot bypass on every tool, plus crawl and batch jobs for many URLs. Source

Tools it exposes · source
scrapeextractcrawl_createbatch_createbrowser_navigateaccount_usage
Free tier or trial
5,000 credits for the first billing period when the local server runs without an API key
Billing
Server is free; each tool call is billed at the underlying product's rate (scrape: 1 to 25 credits)
Our provider rating
4.3/5
Try ZenRows

Sponsored link. Ratings are independent.

Official MCP docs →

"Our provider rating" comes from our review of each provider's core proxy and scraping products. It is not a score for the MCP server.

Setup

Connect each one to Claude Desktop

Add the block to claude_desktop_config.json, then restart Claude Desktop.

claude_desktop_config.json 
{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "<your API token>"
      }
    }
  }
}
Copied from the Bright Data MCP docs. Check them for the current version before you paste.

Which one fits your use case?

If you need…Look atWhy
Search results from several engines for an agentBright Datasearch_engine covers Google, Bing and Yandex.
Amazon search and product data as JSONOxylabsDedicated amazon_search_scraper and amazon_product_scraper tools with optional structured parsing.
Ready-made scrapers for many named sitesBright Datasearch_scrapers and run_scraper work with pre-built scrapers for platforms such as Amazon, Booking.com, LinkedIn and Zillow.
Protected, JavaScript-heavy pagesZenRowsscrape adds js_render and premium_proxy, with anti-bot bypass on every tool.
Thousands of URLs as background jobsZenRowsbatch_* tools submit and manage asynchronous jobs across many URLs.
Describe a task and let an agent drive the browserOxylabsai_browser_agent navigates, clicks and fills forms from a prompt.

FAQ

What is an MCP server?

An MCP (Model Context Protocol) server exposes tools that AI apps can call. Bright Data's documentation describes its server as one that lets LLMs, AI agents and applications access, discover and extract web data in real time. The agent decides when to call a tool such as a page scrape or a search, and the server returns the result.

Do I still need proxies if I use an MCP server?

Not separately. Each server runs on its provider's own infrastructure: Bright Data's docs mention a global IP network and unlocker technology, Oxylabs lists rotating IPs covering 195+ countries, and ZenRows puts anti-bot bypass on every tool. Usage is billed against that provider's account.

Which providers offer a free tier for their MCP server?

All three, per their docs. Bright Data includes 5,000 requests a month, shared with its other products. Oxylabs offers a 1-week Web Scraper API trial and 1,000 free AI Studio credits. ZenRows gives 5,000 credits for the first billing period when its local server runs without an API key.

Can I use an MCP server outside Claude, e.g. in Cursor or my own agent?

Yes. Bright Data lists Cursor, Windsurf, LangChain and custom agents; Oxylabs lists Claude Code, Cursor and any client that supports remote MCP servers with custom headers; ZenRows lists Cursor, Windsurf, VS Code, Zed and JetBrains IDEs.

Advertisement
Explainer What is an MCP server? What an MCP server is, how hosts, clients and servers fit together, local vs remote servers, and how scraping providers use MCP to give AI agents web data. Provider guide Bright Data MCP for agents Bright Data's MCP server from its docs: 32 tools in groups, the 5,000-request free tier, remote and local setup, and settings that limit cost and context. Provider guide Oxylabs AI Studio MCP Oxylabs' MCP servers from their docs: Web Scraper API and AI Studio tools, the AI Studio-only package, Claude setup, and the free trial and credits. Provider guide ZenRows MCP ZenRows' MCP server from its docs: scrape, extract, crawl, batch and browser tools, credits per scrape, remote and local setup, and the free credits. Provider guide Proxies for LLM training data IPRoyal and DataImpulse for AI training data, from their own pages: proxy types, pool sizes, prices per GB, targeting, sessions and what's missing.

Sources