# Disabling subagents seems to have helped with my usage limits — retained record

> Retained reddit record from reddit:ClaudeCode.

Canonical: https://highsignal.app/data/records/3f276fb723b86b82

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeCode)  published 2026-09-19 19:07 ingested 2026-09-20 02:38

[original source ↗](https://www.reddit.com/r/ClaudeCode/comments/1wkun6p/disabling_subagents_seems_to_have_helped_with_my/)

## retained text



```
I’ve been struggling with limits for about a month. After reading a post about subagents consuming more tokens, I asked for a usage breakdown from what felt like a pretty normal session: \*\*111 million input tokens\*\*, roughly \*\*98% cached\*\* (which still counts), and about \*\*260k output tokens\*\*.

Three days ago, I told it to stop using subagents and work sequentially. I also told it to run tests only when I ask, because I noticed it was testing practically every step.

Since then, things have felt closer to how they used to. Fable and Astra still consume more than I’d like, it feels like they burn 3% just for thinking of using them. But overall, it’s much less extreme now. I’ve made more progress than I was managing over the past few weeks.

If I want Fable to plan and Opus to implement—or Astra and Sol—I have the planner write a handoff, then run the implementation in a separate session myself. I control the handoff instead of letting the planner keep directing another agent.

My suspicion is that more aggressive subagent use and constant testing are behind at least some of the increased consumption, intentional or not.

Anyone else tried this and noticed a difference?
```



document fetched 2026-09-20 00:36 · canonical https://reddit.com/r/ClaudeCode/comments/1wkun6p/disabling_subagents_seems_to_have_helped_with_my
