Multi-Agent System

A multi-agent system is an architecture in which multiple AI agents collaborate or compete to complete tasks that are too complex for a single agent.

Agents may act as orchestrators, specialists, critics, or executors, passing context, delegating sub-tasks, and aggregating results. Multi-agent systems require governance at both the individual agent level and at the interaction layer, where data flows, permission grants, and action chains cross agent boundaries.

Closely related concepts include Orchestrator Agent, Multi-Agent Collaboration, Agent Mesh, Delegation Control. Each is defined in the Prefactor glossary.