Run Claude Agent Swarms Without Burning Your Limit
Dynamic workflows run many Claude agents at once, and every one draws on your plan's limit. Three controls that keep the swarm cheap, the one new setting to test first, and what a smaller model is worse at
A dynamic workflow runs dozens of Claude agents from one script. Anthropic's docs say those runs count toward your plan's usage and rate limits, and an agent with no other model assigned runs on your session's model.
Here are the three controls, the one new setting to test before you trust it, and the catches.
What a swarm costs you
- Every workflow agent draws on the same plan limit as your normal chat.
- With nothing else set, each agent runs on your session's model, so an Opus session means an Opus swarm.
- Claude Code warns you ('Large workflow') when a run schedules more than 25 agents or projects over 1.5 million tokens. It is a warning only; it doesn't stop the run.
Control 1: use a smaller model for easy stages
- Run /model before a big run, and switch to a smaller model if that's what you use for routine work.
- When you describe the task, say which stages are easy: 'use Haiku 5.5 for the file-reading and listing stages, keep the strongest model for the final review.'
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (4 more sections, 1 copy-paste prompt) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
Unlocks: Control 2: set the workflow size to small · Control 3: one model for every workflow agent (new) · The catches · Sources
50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.