I'm excited to welcome you to this course on building Interactive Agents with generative UI. Built in partnership with Kilpatrick. Building a capable agent is hard work, but once you've done that, you get a new challenge getting it to users in a way that they enjoy interacting with it, which may require much more than building yet another chat bot interface. Think about what a good agent interface might look like. Instead of outputting long paragraphs of text. Maybe the user would rather just give a simple form to felt or chat to interact with. The agent doesn't just answer in plain text and give the user something to see or act on. In this course, you learn to build that interface using three approaches to generative UI where the UI, the user interfaces generate it. The first approach is controlled components, where you define exactly what the agent can show. The second is declarative UI with the framework a2 UI, which is an open spec originated by Google and co-developed with Copilot, where the agent generates layouts from a schema. The third is open ended UI with MCP apps, where the agent surface is full interactive experiences like whiteboards and approval flows directly inside your product. Your instructor is a tie. The guy who is co-founder of copilot, the team behind the AG UI protocol that this course is built on. Thank you Andrew. That's exactly the problem we built. Copilot gets to solve every interaction between people and technology is increasingly mediated by agents. And yet, as we film this course, most agents still talk to users in plaintext, which is a lot like the command line and mS-DOS era of personal computing. Powerful for early adopters, but not yet built for mass adoption. The Windows and Mac era changed that and brought full featured applications that anybody could use. AI is going through the same transition now and all UI is becoming AI. When a user asks for a report, the agent should build it. When they need to approve something, the agent should show them a form. When they need to explore data, the agent should render a chart. You'll start by connecting a lane change agent to a react front end using Copilot kits and ag UI. From there, you'll build across the full generative UI spectrum from controlled generative UI with handcrafted components that render demand to declarative generative UI, where the agent assembles UI from a component catalog of building blocks to open ended generative UI with app experiences powered first by MCP apps and then by full app sandboxes. By the end of this course, you had built a full stack agent app that renders charts and cards on demand, surfaces, a live wipe through MCP apps, and keeps a share. To do this, in sync between the agent and the UI. Copilot Kit is open source and supports all three of the major generative UI patterns in one framework. It also has first party integrations with line Graph, Google Cloud, AWS, Microsoft Azure, and others. Many people have worked to create this course. I'd like to thank from copilot Kit Hoodie by Kai and Tyler Slayton from DeepMind. AI. Summer Rae also worked on this course in lesson one. You walk through the three major pillars of the generative UI spectrum side by side. What makes each one different and when you use each one? By the end, you'll have the mental model for everything you're going to build in this course. So let's jump in and get interactive.