Catch expired browser agent logins before an overnight run fails

Build me an internal app my ops team opens to check whether the portal logins our browser agents depend on are still working. An expired cookie is the single most common reason an overnight automation fails silently, and right now nobody finds out until a run comes back empty.

The main screen is a list of every saved Browser Use Cloud profile, loaded with List Profiles. For each profile call Get Profile and show its name, the sites it holds cookies for (the stored cookie domains), and when it was last used. Flag anything last used more than seven days ago as stale, since Browser Use's own guidance is to refresh profiles older than seven days to keep login state fresh. Sort broken and stale profiles to the top, because that is the only reason somebody opens this app.

Each row also needs context that Browser Use does not store, so keep it in the app and let people edit it inline: which team owns the login, which Slack channel to notify, which automations break when it expires, and the URL of the account page a check should open.

Give every row a Check button. It fires a cheap verification task with Run Agent Task against that profile, with a prompt along the lines of "open the account page and tell me whether you are still signed in, and say which account name you can see". Then poll Get Session until the session leaves running, and show a green or red result with the agent's own words quoted verbatim plus the time of the check. Save the last result on the row so the list still shows it the next time somebody opens the app. Keep the task to one page load and one question so a check stays cheap.

When a login is broken, a Fix button starts a stealth cloud browser with Create Browser Session bound to that profile, then reads the live view URL from Get Browser Session and embeds it in the app so the person can sign in by hand, clear any two factor prompt, and land on the logged in page. Under the embed put one clear button that means "I am signed in, save it", and that button must end the session properly with Stop Browser Session, plus Stop Session if an agent session is attached. This step is load bearing: profile state, cookies and local storage, is only saved when the session ends, so if the window is abandoned or times out the fresh login is silently thrown away. Say that plainly in the UI, warn before anyone navigates away with a session open, show how long is left before the session times out, and once the session is stopped re-run the check automatically so the row turns green on its own.

Round it out with management actions. Rename a profile with Update Profile. Retire one with Delete Profile behind a confirmation that spells out that the stored login state goes with it. Onboard a new portal with Create Profile, and drop the person straight into the same manual sign in flow afterwards so the new profile actually ends up holding a login.

On any red profile, a Tell the team button sends a Slack message with the slackbot Send a Message action to the channel stored on that profile. The message should name the login that went bad, the sites it covers, what the agent reported in its own words, and which automations are blocked until it is fixed. Pre-fill the text and let the person edit it before sending.

Design it as a dense operational table rather than a pretty dashboard: red, amber and green status, everything visible without scrolling sideways, and quiet when all the logins are healthy.