Multi-agent

A multi-agent system is an architecture where several specialized AI agents collaborate on tasks too complex for any single one. Each agent has its own role, goals, and tools, but coordinates through a defined protocol — delegation, messaging, or a shared state graph. Frameworks that implement this include LangGraph, AutoGen, and Semantic Kernel. The approach shines when a problem naturally decomposes (researcher + coder + reviewer) and when you want parallelism or fault-tolerance from specialization.