Multi-agent
What is Multi-Agent System?
A multi-agent system is a designed workflow with more than one agent role.
Roles need contracts. Extra names without artifacts are theater.
Visual
Multi-agent collaboration
Specialist agents exchange typed artifacts through an orchestrator.
Why it matters
A multi-agent system uses multiple specialized agents that collaborate through handoffs. Each agent has a distinct role and produces a specific artifact type. The system works when roles are clear and handoffs are typed.
Multi-agent systems add complexity. They are justified only when the task genuinely requires different specialist roles that cannot be served by one agent with more tools.
When to use multi-agent
Use multi-agent when: different parts of the task produce different artifact types (code, research, design), when context windows are too small for one agent, or when you need parallel execution.
Do not use multi-agent when: adding more tools to a single agent would work, when the "agents" are just renamed prompt variations, or when handoff overhead exceeds the benefit.
Key takeaways
- 1Multi-agent systems need typed handoffs and clear role boundaries.
- 2Complexity is the cost — only justified when single-agent fails.
- 3Extra agent names without distinct artifacts are theater.
Common mistakes
- ✕Using multi-agent for prestige when a single agent with tools would work.
- ✕Not defining clear handoff contracts between agents.
- ✕Creating "agents" that are just renamed prompt variations.
Related terms
Concept neighborhood
Terms linked from Multi-Agent System in the glossary graph.
- Multi-Agent System
- Orchestrator
- Handoff
- Supervisor Agent