Grok Bot computer freezes/reboots repeatedly and kills 24/7 background scripts

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Grok Bot

Describe the Bug

My Grok Bot runs a 24/7 pipeline on its own computer: bash scripts collect data every 30 minutes and wake a webhook routine to publish. The bot computer keeps freezing for long periods, and once it rebooted. While it is frozen, all background processes stop, so scheduled ticks are missed and data is lost. After the reboot all user processes were gone and did not start again on their own.

Observed (Asia/Yekaterinburg, UTC+5), Oct 9–10, 2026:

  • Freezes: ~20:44–22:01, 22:32–22:46, 22:49–01:03, 01:20–07:38, 08:00–11:12, 11:48–14:21, 19:31–21:25
  • Reboot: Oct 10 ~16:53, background scripts killed

Steps to Reproduce

  1. On the Grok Bot computer, start long-running background scripts (nohup bash loops with sleep, logging a line every 30 min).
  2. Leave the bot idle for an hour or more.
  3. Check the logs: there are gaps of 15 minutes to 6 hours with no entries. The processes are still listed afterwards but did not run during the gap. Sometimes the computer reboots and all user processes are gone.

Expected Behavior

The bot computer stays awake while user background processes are running. Or there is a supported keep-alive / scheduled-wake option, plus a way to autostart user scripts after a reboot (there is no cron or systemd on the box).

Operating System

MacOS

Version Information

Grok Bot: Version 0.20.0, Release Track: stable, OS: darwin.

Additional Information

Workarounds tried:

  • An internal keepalive script freezes together with the computer, so it cannot wake it.
  • Autostart after a reboot is impossible: no cron or systemd.
  • An external scheduled routine as a wake-up works, but every run costs tokens.

Second issue: an Auto-review allow rule for this exact publishing workflow does not always apply. Twice a background publish run stopped at an approval card that expired unanswered, and posts were lost or half-published.

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

7

Hey @TG_Promo, thanks for the detailed report and the timestamps.

What you’re seeing is the computer going to sleep. A few minutes after the app disconnects and no bot is working on it, the computer is paused. Processes running inside it (like nohup loops) don’t count as activity, so they stay frozen until the next wake. That is why the gaps run from minutes to hours. Your freeze windows line up with that. This is a known issue for long-running jobs, and I’ve added your report to it.

The restart on Oct 10 was an automatic computer update. Files carry over through an update, but running processes don’t, which is why your scripts were gone.

For now, a few things help:

  1. Keep the Grok Bot app open and connected while the pipeline needs to run. The computer woke each time your app reconnected.
  2. Have each collector run catch up on the time since its last successful run (keep a timestamp in a file), so a missed tick doesn’t lose data.
  3. Have your publishing routine check whether your scripts are running and start them if they aren’t, so they come back after an update.

On the approval cards: the Auto-review safety check can fail to complete, and when that happens it asks for approval even if an allow rule matches. That part is a separate issue we’re tracking.

8

3 days since i cant use the grok bot for this bug. team didnt reply, days pass and no reply from team.

when will be fixed? i am paying for a suscription that dont work well i cant use!

10

The issue still hasn’t been fixed. My Grok bot periodically goes to sleep, and scheduled tasks fail to run. I set up an extra routine that consumes the bot’s limit, but it at least helps wake the bot up. However, I end up using far more tokens… It’s frustrating.

12

yes , frustrating