AI Agents and Multi-Agent Systems · Lesson 5
Intro and How to Build Multi-Agent Systems (MAS)
A multi-agent system uses several specialised agents — for example a researcher, a writer and a reviewer — that work together on a goal.
Topologies include supervisor (a manager agent delegates and collects results), hierarchical (managers of managers), and network or swarm (agents hand off to each other directly).
To build one: define roles and tools per agent, choose a topology, decide the shared state or messages, and add stopping conditions. Frameworks such as LangGraph, CrewAI and AutoGen help.
Key points
- Specialised agents with clear roles
- Supervisor, hierarchical or network topologies
- Define shared state and hand-offs
- Add stopping rules to avoid endless chatter