A multi-agent system coordinates several specialized AI agents on a shared objective: a planner decomposing work, specialists executing subtasks, reviewers chec
A system that coordinates several specialized AI agents on a shared objective: a planner decomposing work, specialists executing subtasks, and reviewers checking outputs, with handoffs and shared memory between them.
When specialization, parallelism, or independent review measurably beats a single capable agent. It costs coordination complexity and multiplied token spend, so production systems are always wrapped in tracing so humans can audit who did what and why.