Hey, yeah, this is a pretty well known pain on the old Pro plan with request based billing. Each subagent call counts as a separate request, so one chat can easily burn 4 to 6+ requests. That’s just how it works on the legacy plan, but I get why it’s annoying.
Here are a few ways to reduce subagent usage:
- Add a project rule at
.cursor/rules/no-subagents.mdor User Rules in Settings, something like “Don’t run subagents. Do all work in the main agent context.” It’s not a hard block, but the model usually follows it. - Lock to a single model: in the chat model dropdown, turn off Auto and Use multiple models, then pick one model explicitly.
- Turn off composer-2 in Settings > Models. Keep in mind this can sometimes cause tool failed errors when the agent still tries to delegate, so option 1 is the cleaner fix.
Related threads with more community tips: I don't want to use subagents anymore, How can I prevent Cursor from using composer - 2 - fast to create sub - agents?
If anything changes, I’ll post an update in the thread.