WHY USE MORE THAN ONE AGENT

CHAPTER SUMMARY A single-agent system uses one loop and one context to do a whole task, while a multi-agent system uses several coordinating agents. Single agents are simpler, cheaper, and easier to debug, so they are the right default. Multi-agent systems help when tasks need different specialized skills or mindsets, and when separating contexts keeps each agent focused. Specialization lets each agent have a prompt, tools, and permissions tuned to one job. Context isolation keeps clutter from one subtask out of another agent's working memory, improving quality, cost, and sometimes safety. Choose multi-agent design deliberately and with evidence, never by default. KEY TERMS REVIEW Single-agent system: one agent and one loop handling a task end to end. Multi-agent system: two or more coordinating agents, each with its own loop and context. Specialization: giving an agent a narrow job and tuning its prompt, tools, and permissions to that job. Context isolation: giving each agent its own context window so clutter does not spread. Token: a chunk of text that models read and write, and that providers use to measure cost. System prompt: the standing instructions given to a model at the start of a run.