agent-browser

Open a URL with agent-browser open, run snapshot -i to get element refs like @e1, then use click, fill, select, or scroll with those refs. Re-snapshot after any navigation or DOM change because refs are invalidated when the page changes.

Navigate to the form, snapshot with -i to discover field refs, then use agent-browser fill for text inputs, select for dropdowns, and check for checkboxes. Click the submit button and wait with --load networkidle to confirm the result.

Yes. After logging in, run agent-browser state save to persist cookies and storage, then state load in later runs. Alternatively use the auth vault with auth save and auth login so passwords stay encrypted and never appear in prompts.

Yes. Pass --session <name> to run isolated sessions with separate cookies, storage, and tabs, which is useful for concurrent scraping or A/B comparisons. Use session list to inspect active sessions and close each one when finished.

Refs are invalidated whenever the page changes, such as after clicks that navigate, form submissions, or dynamic content loading. Run snapshot -i again to obtain fresh refs before interacting further.

Yes. Use screenshot with --full or --annotate for images, pdf to save the page as PDF, and record start/stop to capture WebM video of the session. The diff command also compares snapshots or screenshots against baselines.