Multi-Agent System
A system where multiple specialized AI agents collaborate to solve a problem each with its own role and tools.
Multi-agent systems split a complex job across specialized agents that communicate. A research crew might have a planner, several specialists, a fact-checker, and a synthesizer. Each one is simpler than a single all-purpose agent, but together they handle work that no single agent could.
Frameworks like CrewAI, AutoGen, and LangGraph all support this pattern. The orchestration is the hard part: how do agents pass results, how do they handle disagreement, how do they avoid burning tokens on useless loops?
Multi-agent isn't always better than single-agent. For most tasks, one good prompt to one good model wins. Reach for multi-agent when the work has clearly separable steps that benefit from specialized reasoning.