Webscraping AI MCP Server

For a product page, the fields tool accepts instructions for attributes such as title, price, and description. The question tool instead answers queries about page content. HTML and text tools retrieve rendered markup or visible text, while CSS-selector tools target either an individual element or several selected elements. An account tool reports usage information alongside these extraction operations, so the MCP client can also check the associated WebScraping.AI account.

Scraping requests can control JavaScript rendering, wait for a matching element, execute custom JavaScript, and emulate desktop, mobile, or tablet devices. Proxy options include datacenter, residential, and stealth services with country selection, as well as a custom proxy. The server manages request concurrency, applies backoff for rate limits, and retries transient errors. A WebScraping.AI API key is required, and configuration examples cover Cursor and Claude Desktop. Optional content sandboxing is disabled by default; when enabled, it surrounds scraped material with boundaries identifying it as external content rather than instructions.