Multi-Agent Systems: When One Agent Is Not Enough
Orchestrator–worker, pipeline and debate patterns, what they cost, and how to decide whether you need more than one agent.
Agentic AIAdvanced2 min readUpdated 11 Oct 2026
Start with one agent
A single agent with good tools and clear instructions solves most tasks. Multiple agents add latency, cost, coordination failures and harder debugging. Add them only when they solve a specific problem.
Good reasons for multiple agents
| Reason | Example |
|---|---|
| Independent sub-tasks in parallel | Research ten companies at once |
| Context isolation | Each worker reads its own large documents; the orchestrator sees summaries |
| Different tools or permissions | A read-only research agent and a separate agent that can write to a system |
| Specialised instructions | A reviewer agent that checks another agent's output |
Common patterns
Orchestrator–worker: orchestrator → plan → [worker A, worker B, worker C] → combine
Pipeline: extract agent → validate agent → summarise agent
Generator–critic: writer → reviewer → writer (until the checks pass or a limit is hit)Make it reliable
- Pass structured handoffs (JSON with fields), not free-form chat.
- Give each agent a narrow role, its own tools and a step limit.
- Keep shared state in code or a database, not only in prompts.
- Log every step with inputs, outputs and token cost so failures can be traced.
Finished this note?