The ZenRows MCP server provides AI systems with real-time access to any website through a scrape tool and 30+ browser_* automation tools, bypassing anti-bot measures to extract content from any webpage.
Scrape any webpage: Fetch content in Markdown (default), plain text, HTML, or PDF formats.
Handle dynamic content: Enable JavaScript rendering (js_render) for SPAs (React, Vue, Angular).
Bypass anti-bot protections: Use premium residential proxies and managed proxy rotation to access protected sites.
Geo-targeted scraping: Route requests through a specific country via proxy_country (ISO 3166-1 alpha-2 codes).
Structured data extraction: Auto-parse products, articles, and listings (autoparse), or extract specific elements using CSS selectors (css_extractor).
Targeted data outputs: Extract specific content types (emails, headings, links, images, videos, etc.) as structured JSON.
Wait/timing controls: Wait a set number of milliseconds (wait) or until a CSS selector appears (wait_for) before capturing content.
Browser interactions: Execute custom actions before scraping (clicks, form fills, waits) via js_instructions.
Screenshots: Capture above-the-fold (screenshot), full-page (screenshot_fullpage), or element-specific (screenshot_selector) screenshots.
Full browser automation: Navigate, click, fill forms, execute JavaScript, and manage cookies, tabs, and sessions across 30+ dedicated browser tools.
The Zenrows MCP (Model Context Protocol) server is the standard way AI systems use Zenrows' web data infrastructure. A single connection gives your AI assistant, agent, or application reliable, real-time access to the live web, including the protected web.
📚 Full documentation: docs.zenrows.com/mcp/overview
Reach sites that normally block bots. Get reliable access to protected sites at scale, without building anti-bot handling yourself.
Managed web data infrastructure. Proxy rotation, headless browser orchestration, anti-bot handling, and session management run on Zenrows' infrastructure.
Plug into any AI you already use. Works with any MCP client, including AI assistants, agent frameworks, AI SDKs, IDE plugins, and custom applications.
Plain English, no scraping code. Describe the task naturally and the AI picks the right tool. No selectors, no proxy management, no anti-bot tuning.
Related MCP server: defuddle-mcp
Zenrows MCP supports two transport options. Both expose the same set of tools and capabilities. Pick the one that fits your client.
Use the hosted Zenrows MCP server when your AI application calls an LLM API directly. The server runs on Zenrows' infrastructure, so there is nothing to install, configure, or update.
Server URL:
https://mcp.zenrows.com/mcpTransport: Streamable HTTP
Authentication: OAuth or API key as Bearer token. Pass your Zenrows API key in the Authorization header on every request (or complete OAuth in clients that support it).
Authorization: Bearer YOUR_ZENROWS_API_KEYMost MCP clients accept this through an authorization shorthand field on the tool config and forward it as the Bearer token automatically. Some clients use a free-form headers field instead. Either approach works.
Remote MCP does not auto-create accounts. Use OAuth “Create Free account” in the client, or pass an existing API key.
import os
from openai import OpenAI
ZENROWS_API_KEY = os.environ["ZENROWS_API_KEY"]
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
response = client.responses.create(
model="gpt-5",
tools=[
{
"type": "mcp",
"server_label": "zenrows",
"server_description": "Web scraping MCP server for accessing live web content.",
"server_url": "https://mcp.zenrows.com/mcp",
"authorization": ZENROWS_API_KEY,
"require_approval": "never",
}
],
input="Visit https://news.ycombinator.com/ and summarize the three most recent posts.",
)
print(response.output_text)For the full walkthrough with framework-specific examples, see the Remote MCP server docs.
Use the local stdio configuration when your MCP client runs the server as a local subprocess instead of calling a remote URL. This is the standard setup for desktop AI tools and IDE plugins, including Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, Zed, and JetBrains IDEs.
Package: @zenrows/mcp on npm
Authentication:
ZENROWS_API_KEY environment variable, or
Key previously stored in ~/.zenrows/secrets.json, or
Auto-signup (default): if neither is set, stdio provisions a Free plan account via POST /api/agent/signup, persists the key + claim metadata under ~/.zenrows/ (secrets.json + account.json, mode 0600), and prints a claim URL on stderr. Opt out with ZENROWS_AUTO_SIGNUP=false.
Requirements: Node.js installed (for npx to work).
Configuration (with your own key):
{
"mcpServers": {
"zenrows": {
"command": "npx",
"args": ["-y", "@zenrows/mcp"],
"env": {
"ZENROWS_API_KEY": "YOUR_ZENROWS_API_KEY"
}
}
}
}Zero-config (auto-signup):
{
"mcpServers": {
"zenrows": {
"command": "npx",
"args": ["-y", "@zenrows/mcp"]
}
}
}The exact location of this config varies by client. See the per-client setup guides for the file path for your client.
The Zenrows MCP exposes these tool families:
Tool | Purpose |
| Full-page content → Markdown, plain text, HTML, PDF, or screenshot (plus helper outputs). |
| Structured JSON ( |
| Cloud Batch API fan-out ( |
| 30+ tools for full browser automation (navigation, clicks, forms, JS, cookies, tabs, sessions). |
The AI selects the right tool from your prompt. You don't call tools directly in code.
See the full tool reference for every tool, parameter, and return value.
git clone https://github.com/ZenRows/zenrows-mcp
cd zenrows-mcp
npm install
cp .env.example .env # Optional: add your API key (stdio can auto-signup)
npm run dev # Run with .env loaded (requires Node.js 20.6+)
npm run build # Compile to dist/
npm run inspect # Open the MCP inspector UIPull requests and issues are welcome.
Scrape any webpage and return its content using ZenRows.
Use this tool to fetch webpage content for analysis. By default it returns clean markdown, which is ideal for LLM processing.
When to enable options:
js_render: page uses React/Vue/Angular, loads content dynamically, or content appears missing on the first attempt
premium_proxy: site returns 403/blocked errors even with js_render enabled
wait_for: specific content loads after initial render (requires js_render)
css_extractor: you only need specific elements, not the whole page
autoparse: structured data pages like products or articles
Examples: Basic: { url: "https://example.com" } Dynamic: { url: "https://spa.com", js_render: true } Protected:{ url: "https://protected.com", js_render: true, premium_proxy: true } Extract: { url: "https://shop.com", css_extractor: '{"title":"h1","price":".price"}' }
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | The webpage URL to scrape | |
| js_render | No | Enable JavaScript rendering via headless browser. Required for SPAs (React, Vue, Angular) and pages that load content dynamically. | |
| premium_proxy | No | Use premium residential proxies to bypass anti-bot protection. Required for heavily protected sites. Implies higher credit cost. | |
| proxy_country | No | Country for geo-targeted scraping. ISO 3166-1 alpha-2 code (e.g. 'US', 'GB', 'DE'). Requires premium_proxy=true. | |
| response_type | No | Output format. 'markdown' (default) preserves structure and is ideal for LLMs. 'plaintext' strips all formatting for pure text extraction. 'pdf' returns a PDF of the page. 'html' returns the raw HTML source (omits the response_type param; ZenRows default). Ignored when autoparse, css_extractor, outputs, or screenshot params are set. | markdown |
| autoparse | No | Automatically extract structured data from the page into JSON. Best for product pages, articles, and listings. | |
| css_extractor | No | Extract specific elements using CSS selectors. JSON object mapping names to selectors, e.g. '{"title":"h1","price":".price-tag"}'. Returns JSON instead of full page content. | |
| wait_for | No | CSS selector to wait for before capturing. Use when key content loads after the initial page render. Requires js_render=true. | |
| wait | No | Milliseconds to wait after page load before capturing content. Max 30000 (30s). Requires js_render=true. | |
| js_instructions | No | JSON array of browser interactions to run before scraping. Requires js_render=true. Example: [{"click":"#load-more"},{"wait":1000},{"wait_for":".results"}] | |
| outputs | No | Comma-separated list of data types to extract as structured JSON. Available: emails, headings, links, menus, images, videos, audios. Use '*' for all types. Returns JSON instead of full page content. | |
| screenshot | No | Capture an above-the-fold screenshot of the page. Returns an image instead of text content. Useful for visual verification or debugging. | |
| screenshot_fullpage | No | Capture a full-page screenshot including content below the fold. Returns an image instead of text content. | |
| screenshot_selector | No | Capture a screenshot of a specific element using a CSS selector. Example: ".product-card". Returns an image instead of text content. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and destructiveHint=false. Description adds valuable behavioral context: default markdown output ideal for LLMs, and crucially explains that certain parameters (css_extractor, autoparse, outputs, screenshot) change the return type from text to JSON or images. This output-switching behavior is not captured in annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Well-structured with clear information hierarchy: purpose statement, default behavior, conditional options guide, and examples. Every section earns its place. Examples section is slightly verbose but appropriate for a 14-parameter tool where syntax matters. Good use of formatting (bullet points, code blocks).
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a complex tool with 14 parameters and no output schema, description adequately explains return value variations (markdown default vs JSON vs images depending on params). Covers the ZenRows-specific options (premium_proxy credit cost mentioned in schema, wait_for interactions explained). Could mention error handling or rate limits, but sufficient for invocation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, establishing baseline 3. Description adds significant value via the 'When to enable options' section which provides contextual semantics for when to use parameters (e.g., 'page uses React/Vue/Angular' triggers js_render). The concrete examples demonstrate parameter interactions and valid value formats (e.g., CSS selector JSON syntax).
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Opens with specific verb+resource ('Scrape any webpage') and identifies the underlying service ('using ZenRows'). Clearly states default output format ('clean markdown') and primary use case ('fetch webpage content for analysis'). No siblings to differentiate from, but scope is precisely defined.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Contains explicit 'When to enable options' section that maps specific technical conditions (React/Vue/Angular, 403 errors, delayed content loading) to parameter usage. Provides concrete decision trees for selecting js_render, premium_proxy, and other options. Includes practical JSON examples showing parameter combinations.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
With only one tool, there is no possibility of confusion or overlap between tools. The single 'scrape' tool has a clear, distinct purpose of fetching webpage content.
There is only one tool name, so consistency is inherently perfect. The name 'scrape' follows a clear verb-based pattern appropriate for its function.
A single tool is too few for most server purposes, as it limits functionality and flexibility. While scraping is a focused domain, having only one tool feels thin and may not cover related needs like batch processing or error handling.
The tool covers basic webpage scraping with options for dynamic content and proxies, but there are notable gaps. Missing operations might include checking scrape status, managing sessions, or handling rate limits, which could lead to agent workarounds or failures in complex scenarios.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
One MCP for the Web. Easily search, crawl, navigate, and extract websites without getting blocked.…
Scrapingdog MCP — wraps Scrapingdog (scrapingdog.com), a proxy-based web
Firecrawl MCP — wraps the Firecrawl API (firecrawl.dev) for web
Turn any URL into clean Markdown and structured data. Scrape, crawl, search and extract.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ZenRows/zenrows-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server