Coordinator and Specialist Roles
Explore the roles of coordinator and specialist agents within hub-and-spoke multi-agent architectures. Understand how to isolate context, delegate subtasks efficiently, and design tool access to enhance system reliability and avoid context pollution.
In the previous chapter, we introduced multi-agent systems as the right architecture when subtasks are independent. This chapter goes deeper: How do we actually structure the agents, divide the work, and keep the contexts clean? This lesson covers the hub-and-spoke architecture that underpins multi-agent systems and walks through the concrete decisions around what each agent knows, which tools it can access, and how the coordinator and specialists divide responsibility.
By the end of this lesson, we will be able to:
Describe the hub-and-spoke architecture and why it outperforms flat agent arrangements.
Define the coordinator's role and what it must not do.
Define the specialist's role and what context it should and should not receive.
Design a task split for a concrete research or support scenario.
The hub-and-spoke architecture
In a flat agent architecture, one agent handles everything: all tools, all context, and all decisions. As the task grows, so do the tool list and the message array. Tool selection degrades. Context fills up. The agent starts making mistakes that a fresh, focused agent would not. The hub-and-spoke ...