web_fetch returns one public web page rendered in a fresh browser as
Markdown, HTML, plain text or a base64 PNG screenshot, with the status
code the site returned, selectors, wait conditions, page timeout, locale,
user agent, link and image styles, iframes and shadow DOM. It needs an
API key.

num on search and google_search now accepts 1 to 10, the most Google
returns on one page.