Multi-agent systems
Also known as: multi-agent · agent teams
A multi-agent system is several agents working together on a task — one orchestrating, others specializing — the way a small team divides work.
Common patterns:
- Orchestrator + workers: a lead assigns subtasks to specialists.
- Debater: agents argue a question from different sides, a judge decides.
- Pipeline: each agent handles a stage (research → draft → edit).
Don't reach for it reflexively: multi-agent is more moving parts, more cost, more failure modes. Start with one good agent; split only when a single context or role clearly can't hold the job.