The recurring job

A page watch starts with one URL. The next run needs the current page, plus the title and canonical URL that explain where the content came from. Ask for markdown when the result feeds a later agent step.

scrape-website follows redirects and returns the final URL, response status, page metadata, the main h1 heading, and the requested body format. Keep the request and the returned final_url with the watch state. A changed redirect or canonical URL is then visible to the next run. 

curl -X POST https://x402.agentutility.ai/scrape-website \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/updates","format":"markdown","include_links":false}'

The first response requests x402 payment. After payment, save the response beside the watch record. It includes the content to inspect and the metadata needed to decide whether the page still represents the same source.

Keep the watch input small

Store one request body per watched page. Change the URL only when the target changes. Turn on include_links for a link audit, then leave it off for body watches where the body is the useful result. 

{
  "url": "https://example.com/updates",
  "format": "markdown",
  "include_links": false
}
agentutility price scrape-website
agentutility call scrape-website --data "$(cat watch.json)" --max-price 0.04 > page.tmp && mv page.tmp page.json

page.json keeps the final URL, title, canonical URL, and markdown together. That is enough context for a later agent to compare the current result with the prior one.

Choose the right fetch

Use scrape-website for a public static or server-rendered page when the next step needs readable text or metadata. Use website-screenshot when the job needs a rendered image from a JavaScript-heavy page.

The same endpoint is available from @agentutility/mcp-web-probe as scrape-website. Its inputs are url, format, include_links, and user_agent, so an MCP caller can save the same watch settings it uses over HTTP.

Current recovery proof

The 2026-10-29 wallet-recovery snapshot records three repeat buyers for scrape-website. The cohort made 27 calls and spent 1.08 USDC; two recently quiet buyers are recoverable. This page keeps the repeated job focused on one stable URL and a saved response shape, so the next paid call can continue the same watch.