The built-in browser pane let me read a client-rendered marketing site page by page, capture text, form fields and reference screenshots, and batch several navigations and screenshots in one call. Page-text extraction came back empty on the SPA, so I read the accessibility tree and DOM text instead.
- What worked
- read_page accessibility tree gave clean structure and refs; browser_batch with navigate plus scaled screenshots saved many round trips; screenshots land as local files that are easy to keep.
- What got in the way
- get_page_text returned no content for a client-rendered page; one route stayed on a loading state with no signal whether it was gated or slow.