Checkpoint big agent jobs before a usage limit hits
Hit the 5-hour usage limit 30 minutes into a big workflow · X · @georgfranz · Sep 2026
A multi-agent Claude Code workflow stalled at about 90% done when the usage cap hit, followed by a long wait.
The honest numbers
- Says the limit hit after 30 minutes
- Says the work was 90% done
- Says he waited 4h 30m
As stated by the source. Not independently verified.
Lesson
Split big jobs into checkpoints that commit progress, so a usage cap pauses you instead of stranding half-finished work.
Fix prompt
Copy this into your own AI tool (Cursor, Lovable, Claude Code, Bolt…). It's general guidance written by us. Review what the AI suggests before applying it.
I keep running out of usage or credits with Claude Code. Help me work more efficiently without lowering quality. 1) Suggest how to split my current task into small checkpoints I can commit, so a limit pauses me instead of losing work. 2) Tell me which kinds of edits I can do myself or with a cheaper model or mode (plan/chat mode, smaller model). 3) Write a short project brief (stack, key files, rules) I can reuse so I don't spend tokens re-explaining. 4) Give me a checklist to review my plan's usage dashboard and set alerts or a spending cap. Don't start coding yet.
Source: X ↗