The AI agent hype is deafening. The actual understanding of what an agent is: near silence.

Most developers have built a chatbot and called it an agent. They're not the same thing, and the gap matters. An agent perceives, decides, acts, and loops back. Once you see that pattern, you can't unsee it, and you can build one in under an hour.

In this hands-on workshop, we'll build a real agent from scratch, live, using TypeScript, Google's Agent Development Kit and Gemini. We'll start with a single LlmAgent and one tool, then layer the patterns worth knowing: SequentialAgent for chained reasoning and ParallelAgent for fan-out work with an aggregator, rather than all twenty-one patterns that exist. Every architectural decision gets explained as we make it.

You'll leave with working code running on Gemini, a mental model that actually holds up, and a clear path to deploying your agent on Cloud Run.