SURVEY OF AGENT FRAMEWORKS

ROLE-BASED AND CREW-STYLE MULTI-AGENT FRAMEWORKS The second family organizes work around roles, in the way a small company assigns jobs to people. In such a framework, you create several agents, and you give each one a role, a goal, and often a short backstory that shapes how it behaves. You then group them into a crew and give the crew a shared task. For example, a content crew might include a researcher who gathers facts, a writer who drafts text, and an editor who polishes it. You studied the underlying ideas in Part Ten, including the pipeline and supervisor patterns of Chapter 43. A crew-style framework gives those ideas friendly names and ready-made wiring. One widely known example of this family is CrewAI. The main strength is that the mental model is easy to grasp, because everyone understands the idea of team members with jobs. It can also get a prototype working very quickly. The main costs deserve honesty. When the framework handles the hand-offs for you, it can be harder to see exactly what each agent was told and why it did what it did. Hidden prompt text added by the framework can surprise you. And, as Chapter 42 warned, more agents mean more cost and more ways to fail, so a crew should exist only when the task truly needs several roles.