FAQ & troubleshooting

Sessions

A session says “Waiting for you” but I see nothing to approve. Open the session: the approval card is inline. If you truly see nothing, the request likely expired; the run fails with a clear reason and can be continued.

My agent stopped mid-run with “request timed out”. The runtime hit its request limit. Use Continue on the failure card; the agent picks up from its last turn.

Why is the context ring red? The session is near the model’s context window. Compact happens automatically; you can also switch to a bigger-context model for the next turn.

Tasks & schedules

A scheduled task didn’t run while my laptop was off. Eligible schedules catch up when a host is back (within the compensation delay), so it should appear in the task’s run history. If it didn’t, check that the schedule’s timezone and active-period rules actually cover that time.

Two runs fired for one schedule. Compensation runs are marked; a manual re-run plus a catch-up can legitimately double-fire in the same window. Check the run list; triggers are labeled.

Computers & runtimes

My computer shows Offline. The daemon connects outbound; check the machine is awake and the daemon process is running (the desktop app manages it for you). Offline computers keep their history and come back without re-pairing.

A runtime is missing from the picker. The picker lists runtimes detected on paired hosts plus cloud. Install the CLI on the host (its detail page shows the installable catalog) and refresh.

Data & privacy

Where does local execution run? On the paired machine, under your own CLI subscriptions. Sessions stream metadata (status, tokens, results) to the workspace, but execution and files stay local unless a tool you approved sends them somewhere.

Still stuck?

Ask in a session (your agent can usually diagnose its own workspace) or reach us through the site’s contact form.