Troubleshooting

Can My Agent Keep Working While I’m Away?

Yes — but only if you set up background execution on purpose. Here’s how to stop “it stopped when I closed chat” confusion.

A common support loop: someone gives a task, closes chat, comes back later, and assumes the agent failed. In many cases, nothing is broken — the task was never scheduled to run independently.

Bottom line: if work must happen later, use an explicit scheduler (cron/reminder/heartbeat), not a one-off chat request.

What works in the background vs what does not

Use this “offline-safe” task format

Schedule this task to run at [exact time + timezone]. If successful, deliver results to [exact destination]. If it fails, send a short failure summary with next action. Include the word "REMINDER" in the delivery message title.

Pro tip: exact time + exact destination removes most background-task ambiguity.

2-minute verification flow

  1. Create a tiny scheduled task for 5-10 minutes from now.
  2. Specify destination clearly (DM, server channel, or dashboard session).
  3. Wait without interacting.
  4. Confirm both completion and destination correctness.

If it didn’t arrive, check these first

High-friction gotcha: it ran, but in the wrong place

Fast fix: always include explicit destination text like: “Deliver to this exact channel/thread,” and ask for immediate confirmation of target before the job starts.

Known-good test prompt (copy/paste)

Set a reminder for 10 minutes from now. Timezone: America/New_York. Send delivery to: this exact conversation thread only. Message title must start with: REMINDER TEST. Before finishing, confirm: exact fire time, timezone, and exact destination thread/channel.

Escalation packet for #help

Background-work verification issue Task requested: Expected fire time + timezone: Expected destination: Actual result: Was cron/reminder used explicitly? (yes/no) Any error text shown: