ZenRows' MCP server lets AI agents fetch live web pages through ZenRows' scraping and cloud-browser products, with anti-bot bypass on every tool. This page covers the tools, what each call costs and how to connect it, using ZenRows' MCP docs, tool reference and GitHub repo.
What it is
The overview describes a server that gives MCP-compatible tools, agents and apps access to live web pages, combining scraping and browser automation with anti-bot bypass on each tool. It is open source under the MIT license (GitHub).
The tools
The tool reference groups them like this:
| Group | Tools | What they do |
|---|---|---|
| Scrape and extract | scrape, extract, prepare_domain, prepared_domain_status | Fetch one page, or pull structured JSON fields such as a product's name and price |
| Crawl | crawl_create, crawl_wait, crawl_results, crawl_stop and more | Start at one URL and follow the links it finds |
| Batch | batch_create, batch_wait, batch_results, batch_cancel and more | Run asynchronous jobs over many URLs |
| Browser | browser_navigate, browser_click, browser_type, browser_scroll, browser_screenshot and more | Drive a cloud browser session with built-in residential proxies |
| Account | account_usage | Show your plan, credits and usage, free of charge |
The tool reference marks some tools as destructive, such as batch_delete, crawl_stop and browser_evaluate. If your client asks before running tools, those are the ones to watch.
The scrape tool
scrape is the tool an agent will call most. Its main parameters, from the tool reference:
response_type.markdown(the default),plaintext,pdforhtml.js_render. Renders the page in a headless browser.premium_proxy. Uses residential proxies.proxy_country. A two-letter country code, such asUS.wait_forandwait. Wait for a CSS selector, or a set time up to 30,000 ms. Both needjs_render.css_extractor,autoparseandoutputs. Return JSON instead of page text: your own selector map, the main data, or types such asemails,linksorimages.screenshot,screenshot_fullpageandscreenshot_selector. Capture the page or one element.
Setting js_render or premium_proxy turns off Adaptive Stealth Mode (overview). Setting mode_auto to false turns it off as well.
What it costs
The overview says the MCP server is free and each tool call is billed at the rate of the product behind it:
| Call | Cost |
|---|---|
scrape, basic | 1 credit |
scrape with js_render | 5 credits |
scrape with premium_proxy | 10 credits |
scrape with both | 25 credits |
extract | Free for now; billing may apply later |
prepare_domain, prepared_domain_status, account_usage | Free; each new prepared domain uses one monthly slot |
| Crawl, batch and browser calls | Their own product pricing |
Because an agent picks the settings, ask it to try a basic scrape first and only add rendering or premium proxies when a page fails.
Free credits
Run locally without an API key, the server creates a free account on the first tool call. The overview says it gets 5,000 credits for its first billing period, an unclaimed account is not renewed after that, and claiming it keeps the key and usage. Set ZENROWS_AUTO_SIGNUP=false to turn this off (GitHub). The remote server doesn't create accounts.
Connect it remotely
The hosted server is at https://mcp.zenrows.com/mcp over Streamable HTTP. You sign in with OAuth or send your API key as a Bearer token (GitHub). The Claude Desktop guide calls the remote connector the quickest way to use ZenRows in Claude, with nothing to install.
Connect it locally
The local server needs Node.js. For Claude Desktop, add this to claude_desktop_config.json (Claude Desktop guide):
{
"mcpServers": {
"zenrows": {
"command": "npx",
"args": ["-y", "@zenrows/mcp"],
"env": {
"ZENROWS_API_KEY": "<your API key>"
}
}
}
}
The overview lists setup guides for Claude, Claude Code, Cursor, Windsurf, VS Code, Zed and JetBrains IDEs.
Where it fits
ZenRows' server leans toward fetching pages that resist scraping and running jobs over many URLs, with crawl and batch tools built in. It has no search tool, so pair it with another source if your agent needs search results. The comparison of MCP servers shows it next to Bright Data and Oxylabs. For our verdict on ZenRows' core product, see the ZenRows review.
FAQ
How much does a ZenRows MCP call cost?
The server itself is free. A scrape call follows Fetch pricing: 1 credit for a basic request, 5 with js_render, 10 with premium_proxy and 25 with both. extract is free for now, though the docs say billing may apply later. Crawl, batch and browser calls follow their own pricing (overview).
Can I try it without an API key?
Yes, on the local server. Run without ZENROWS_API_KEY and the first tool call creates a free account with 5,000 credits for its first billing period. The docs say an unclaimed account is not renewed after that period. Set ZENROWS_AUTO_SIGNUP=false to turn this off. The remote server does not create accounts.
Can I choose the country a page is fetched from?
Yes. The scrape and extract tools take proxy_country, a two-letter country code such as US. Per the tool reference, it works on its own with Adaptive Stealth Mode; without it, it needs premium_proxy when js_render is set.
Does ZenRows' MCP server include web search?
Not as a tool. The tool reference lists scrape, extract, crawl, batch, browser and account tools, and none of them is a search tool.