Day 1 is your practical introduction to building and shipping a real AI agent in TypeScript. This workshop doesn’t just show you how to wire up an LLM it walks you through many pain points you’ll hit, from authentication to debugging tool calls, and gives you the fixes you need to avoid the classic “black box” trap.

You will:

Choose and configure a modern LLM (like Google Gemini) using the Vercel AI SDK
Wire up your first model call: build a chat backend route, connect it to your frontend with useChat
Set up Discord authentication so only logged-in users can chat, and implement a user-friendly sign-in flow that blocks anonymous access
Upgrade your agent into an actual agent by integrating the Serper search API as a tool
Refactor your frontend to render every tool call, argument, and result using the MessagePart type
Optionally, experiment with native search grounding and external MCP tool integration
Build robust rate limiting for both authenticated and anonymous users

By the end of Day 1, you’ll have a working AI agent that chats, searches, and shows its work with every step, every tool call, and every source. You’ll ship with confidence, knowing exactly what your agent is doing and why. No more black boxes. No more guesswork. Just clear, transparent, debuggable AI, ready for your next big idea.