Module 08 · Agents & Agent Systems

How AI systems plan, act, and coordinate

This module explains the building blocks around a model: agents, tools, loops, memory, permissions, and the point where one agent becomes a team of agents.

Product Manager lens: choose the smallest architecture that can meet the user outcome with acceptable quality, latency, cost, and control.

01AI AgentA system that uses a model, tools, and a control loop to work toward a goal. 02Agent LoopObserve, decide, act, and continue until the task is complete or needs review. 03Tool CallingThe boundary between a model requesting an action and an application executing it. 04Memory & StateInformation that helps an agent track current work and useful history. NEWAgent HarnessThe runtime layer around a model that supplies context, tools, permissions, memory, and controlled execution. 05MCPA connection layer for exposing tools and resources to AI applications. 06Skills / PluginsPackaged capabilities that extend an AI system. 07Human in the LoopReview and approval at the moments where people should retain control. 08 · NEWMulti-Agent SystemsA coordinated team of specialized agents with explicit roles and handoffs.

Open the Module Glossary →