Skip to lesson
Exit
Browser control on a running app1 / 2

1 min lesson

Set the page, session and action boundary

Decide when Browser is needed, then set its page, session and action boundary before it runs.

Step 1 of 2

Browser can reproduce what a signed-in user sees, but its workspace session must be checked before it acts.

Browser is built into Agent and needs no external setup. It can navigate, click, type, scroll and capture screenshots. It can also read console output and, in the Agent panel, network traffic. Browser logs are stored in files so Agent can read only the relevant lines. Cookies, local storage, session storage and IndexedDB persist by workspace, so check the signed-in account before acting. Browser tools ask for approval by default. Keep manual review for sign-in, form submission, unfamiliar sites and any destructive action.

Where this fits
Entry point
@Browser with an exact URL, test account, steps and expected result
Source
Cursor Browser docs

Check the native tools, workspace session behavior and action approvals in Cursor's current Browser docs.