Queue and automation problems
Diagnose waiting work, unavailable inputs, provider failures, and uncertain results.
Inspect the queue or run before restarting it. A failed or cancelled automation may already have changed something in the phone app.
A queue is waiting
Read its waiting reason and inspect its workers.
| Reason | Check |
|---|---|
| Waiting for a ready device | Agent connection, USB, runtime readiness, and whether the device is already assigned elsewhere |
| Waiting for an execution slot | Existing active work and the relevant concurrency limits |
| Waiting for resource capacity | Whether an account or another required resource is held by active work |
| Waiting for access approval or renewal | Your Titan access status |
| Worker paused | Whether that individual worker needs to be resumed |
| Queue paused | Whether the whole queue should be resumed |
Resuming a queue does not automatically resume a worker you paused separately. A soft-paused queue may still have work finishing.
A draft cannot be opened
Drafts are listed in the browser but their detail view is unavailable. Continue preparing the draft through your MCP-connected assistant. Finalize it to Ready, then open the queue and use Play when you intend to start.
A required resource is missing or changed
Have your assistant reread the setup's requirements and the selected resource. Check ownership, row ID, type, schema version, and whether the resource changed after preparation.
For an account already in use, inspect the run holding it. For unavailable media, check its source machine and file availability. Resolve the specific input and review the setup again before starting.
The AI adapter fails
Open AI Adapters and inspect its connection and last test status. Reconnect an expired authorization or correct the key, model, entitlement, or credit issue reported.
If the failure appears only in a supported proxied queue, inspect the selected proxy too. An adapter test uses a direct connection, so it does not prove that the queue's proxy route works.
A start or save result is unconfirmed
Refresh the queue, run, or saved record. Ask the assistant to read the current state before repeating a write. A timeout can occur after a change was accepted.
Check My Runs for an existing execution before creating another. Preserve the ID and message if the outcome remains unclear.
A finished run has failures
Open its run monitor. Review the overall outcome, attempts, final available evidence, and target app. Identify whether a failed attempt was later retried successfully or whether the run ended without its intended result.
Use Re-queue only after checking what the previous attempt did. Hard pause and cancellation do not roll back actions already completed in another app.
A strategy disappeared or access was refused
Check your filters and ask the owner whether the strategy was unpublished, archived, or its grant revoked. Metadata visibility, source review, and permission to execute are separate.
If you need help, collect the queue/run ID, exact message, and relevant attempt for support.