Zenrows MCP

What zenrows-mcp does

zenrows-mcp exposes ZenRows web access as MCP tools for AI assistants, coding agents, and custom applications. It can retrieve a page as Markdown, plain text, HTML, PDF, or a screenshot; return structured fields for products, articles, and listings; submit asynchronous Batch jobs; discover and collect URLs through Crawl; and operate a hosted browser session.

The server is suited to live-web search and data extraction tasks where ordinary HTTP requests may encounter JavaScript-heavy pages, geographic restrictions, or bot protection. ZenRows manages proxy rotation, browser infrastructure, and anti-bot handling behind the API.

How it works

Two transports expose the same capabilities:

• A local stdio process launched with npx -y @zenrows/mcp.
• A hosted Streamable HTTP endpoint at https://mcp.zenrows.com/mcp.

Local use can authenticate with ZENROWS_API_KEY, a key stored under ~/.zenrows/, or the default account auto-signup flow. Remote tool calls require OAuth or a ZenRows API key sent as a Bearer token. The hosted endpoint itself does not create accounts.

Single-page scrape and extract requests use automatic mode by default. ZenRows adds JavaScript rendering or premium proxies only when the page requires them. Supplying js_render or premium_proxy forces that configuration for the request instead. batch_create handles large URL sets asynchronously, while crawl_create starts from one page and follows matching links within the start domain. Batch and crawl progress can be polled or awaited with their corresponding tools.

Setup and configuration

Install Node.js, then add the local server to an MCP client using a configuration similar to:

{
  "mcpServers": {
    "zenrows": {
      "command": "npx",
      "args": ["-y", "@zenrows/mcp"],
      "env": {
        "ZENROWS_API_KEY": "YOUR_ZENROWS_API_KEY"
      }
    }
  }
}

The key can be omitted to allow local auto-signup. Set ZENROWS_AUTO_SIGNUP=false to disable that behavior. ZENROWS_HOME can be used when credentials and account data need to persist in a chosen location, and ZENROWS_TELEMETRY=off omits the machine and host-agent identifiers used during signup. ZENROWS_CRAWL_API_BASE and ZENROWS_BATCH_API_BASE override the default API bases for their respective tool families.

For a remote connection, configure https://mcp.zenrows.com/mcp with OAuth or an Authorization: Bearer token.

Tools and capabilities

• scrape returns full-page content in several formats, including screenshots and PDFs.
• extract returns structured JSON, with automatic or CSS-based extraction options.
• batch_create, batch_status, batch_results, batch_wait, and batch_cancel manage asynchronous URL jobs.
• crawl_create, crawl_get, crawl_results, crawl_wait, crawl_download, crawl_content, crawl_list, and crawl_stop manage link discovery and stored page content.
• browser_navigate and the other browser_* tools support navigation, clicks, typing, form controls, keyboard input, and session cleanup.
• account_usage reports credit allowance and consumption.

Crawl limits such as max_pages and max_items help bound collection. Crawl output can be downloaded as JSON Lines locally, while stored result links are short-lived.

Limitations and notes

ZenRows usage is credit-based: basic requests cost 1 credit, JavaScript rendering costs 5, premium proxy use costs 10, and combining both costs 25. Forced settings can therefore cost more than automatic mode. Batch requires account access, and Crawl must be enabled for the account. Browser sessions hold a concurrency slot until closed. Crawl and Batch jobs are asynchronous, and crawl results may remain partial while a run is active.