Do subagents consume usage?

Subagents still perform model work. Parallel execution is not additional free allowance.

Sources reviewed:

Subagents still perform model work. Parallel execution is not additional free allowance.

Codex

Model and tool work across tasks still draws on the applicable shared allowance.

Claude Code

Subagents work in separate contexts. Their model calls and context contribute to consumption.

When it helps

Delegate independent work and avoid repeating the same background. Parallelism primarily reduces waiting.

Sources & status

OpenAI · Codex usage ↗

Claude Code · Subagents ↗

Back to the tracker →

Copy the address below