Crawlbase MCP Server

Web research and product monitoring in Claude, Cursor, and Windsurf can use Crawlbase MCP to fetch current page content. The crawl tool retrieves HTML from a URL, crawl_markdown extracts Markdown, and crawl_screenshot captures a webpage image. Crawlbase's infrastructure provides JavaScript rendering, proxy rotation, and anti-bot handling for web access. Agents can use the resulting content to examine competitor sites, follow product prices and reviews, or work with news and financial pages.

For larger collections, asynchronous crawling can place results in Crawlbase Cloud Storage for later use rather than an immediate response. Agents can later retrieve saved pages, list stored content, or delete it when the work is finished. This supports revisiting datasets across sessions without putting heavy HTML into the chat immediately. Local client configuration uses stdio with Normal and JavaScript tokens. An HTTP mode supports shared deployments and per-request token headers, with supplied headers taking precedence over environment settings.