Reddit user reports lower Claude usage after changing stuck subagents
A Reddit user said they changed Claude helper agents so an agent that became stuck would terminate instead of retrying indefinitely. The user reported that their daily usage fell from over 30% to less than 10%, while claiming the work was completed with fewer tokens; these observations were not independently verified.
The post offers a practical, user-reported way to reduce token consumption in Claude workflows, but its figures and underlying mechanism remain anecdotal.
See I’m not exactly the brightest but I found some great advice for the less neurologically inclined. Yesterday looked into the /usage for my claude, and for every single helper / subagent it was putting my entire conversation history into it (and mind you, it has 5min cache) so I was putting in 700k+ tokens into double digit agents just for simple changes.
That burns usage!
Today, I made the agents so if one’s stuck it just terminates instead of endlessly trying, burning millions of tokens. My usage dropped from 30%+ over the last few days to less than 10% today. I should be able to stretch out my remaining usage a couple days.
This is literally getting the same stuff done, just lowering token usage drastically. For the non technical doing technical work, this advice may help very much.
For you technical ones, before commenting things remember this is for the NON technical like me. Aka the dull tools in the shed.