You now know how an LLM works under the hood. This module is where that turns into something that does things: an agent. The short version is that an agent is an LLM wired into a loop, given tools, and left to decide for itself what to do next — which is a genuinely different kind of system from a chatbot that just answers you.
It starts by drawing that line clearly: LLMs vs. AI workflows vs. AI agents, and what actually changes as you move from one to the next. From there it gets into the mechanics — the core components every agent is built from, and ReAct (Reason + Act), the reasoning-and-acting loop that underpins almost every agent framework you'll come across. It then widens out to cover tools, memory, and reasoning models together, since that's how they actually show up in a real system. It closes by building a first, tiny agent so the theory has something concrete to attach to.
Watch the lessons in order, then continue on to Prompt Engineering.
The line between an LLM, an AI workflow, and a true AI agent, with real-world examples and a first look at ReAct and RAG — the fastest way to understand what actually makes something "agentic."
The core components every agent is built from — perceiving its environment, reasoning about what to do, and taking action — plus a tour of the different types of agents, from simple reflex agents to more sophisticated goal-driven ones.
The Reason + Act loop at the heart of most agent frameworks: how an agent thinks about what it needs, takes an action to get it, observes the result, and repeats — the pattern that turns an LLM into something that can actually pursue a goal.
A beginner-friendly deep dive tying the pieces together: how an agent differs from a standard LLM, how tools and memory extend what it can do, and how reasoning models change the picture — the fullest picture yet of what's actually happening inside an agent.
A short, hands-on walkthrough of building a real agent with Google's Agent Development Kit — seeing the concepts from this module turn into actual code you can run.