AI Glossary
Multi-agent system
multi-agent system, multi-agent, system of multiple agents
A multi-agent system is a setup in which several specialized AI agents collaborate on a single task, splitting roles and handing results to one another, rather than relying on one general-purpose agent.
- Instead of one general agent, several specialized ones work together.
- The agents split roles and pass on the results of each step.
- It describes the collaboration structure itself, not the layer that controls it.
A multi-agent system is an approach in which a single problem is handled not by one AI agent but by several specialized agents working together. One may plan, another retrieve data, a third write code or verify the result — and together they produce an answer that is more accurate and more robust than with a single general-purpose agent, which quickly loses effectiveness on complex tasks.
The term is sometimes confused with agent orchestration, but they describe different things. A multi-agent system is the structure itself — who the agents are, what roles they hold, and how they are linked. Orchestration is the control layer over that setup: it keeps the order of operations, passes results between agents, and merges them into a single outcome. In other words, a multi-agent system is the "what," and orchestration is the "how we manage it."
In practice, such a system is often ordered by a pre-designed agentic workflow when the steps are predictable, or by more flexible coordination when the path to the goal depends on context. The more agents, the greater the risk of cascading errors, which is why a clear division of roles, control over data flow, and oversight points where a human can step in all matter.
Related terms
Related articles