Skip to main content
Dates are when each change was documented.
Monitors
  • Every monitor now has a thumbnail_url: a 640-pixel-wide screenshot of the first page it watches, taken after you create the monitor and again when an edit changes that page.
Monitors, map, brand and company
  • Search targets for monitors. Watch up to 5 web searches and hear about new results that match your goal. See Targets.
  • Automatic pausing. A monitor whose every page fails for 10 checks in a row, over at least 24 hours, is paused and its recipients are told.
  • /map ranks results most useful first, and the first 10 links carry their page’s title and description. A map that finds nothing fails with NO_LINKS_FOUND, free.
  • /brand returns the color each role uses on the page in brand.colors.roles, and finds social links in more places.
  • /company returns the site’s key pages in links: about, pricing, careers, contact, blog and docs.
Monitors
  • Monitor limits now depend on your plan: shortest schedule, monitors per key, pages per check, history and email recipients. See Limits and credits.
New endpoints
  • /search: web search, optionally with every result page scraped.
  • /transaction: the merchant behind a card transaction line.
  • /styleguide: a site’s design system, with CSS and Tailwind exports.
  • /batch/{endpoint}: run brand, enrich, company, classify, screenshot or styleguide on up to 500 URLs as one job.
  • Logo CDN: any company’s logo as an embeddable image, from a monthly quota instead of credits.
  • /screenshot gains devices, element captures, dark mode, ad blocking and WebP.
Batch scrape and crawl errors
  • /batch/scrape: scrape up to 500 known URLs as one job.
  • GET /crawl/{id}/errors explains a crawl that came back smaller than expected.
  • Every error now carries a stable error_code and a retryable flag. See Error handling.
OAuth
  • Apps such as Claude and ChatGPT can sign you in with OAuth instead of an API key. See Authentication.
Monitors and resolve
  • Monitors: watch pages and sites on a schedule, judge changes against a goal, and get them by webhook or email.
  • /resolve: turn a company or brand name into its domain.
MCP server
  • The MCP server at https://api.prefetch.io/mcp gives Claude, Cursor and other agents every Prefetch tool, with nothing to install.
Web data
  • /scrape, /map and /crawl: any page or site as clean markdown, summaries or structured JSON.