All of you welcome back. In this project video, I’m showing you how to build your first ever multi‑AI agent workflow where one main orchestrator agent distributes work to other agents—research, planning, execution—and then you get a clean final result. This is not theory. I demo the full app UI: a sidebar to configure provider/model/temperature, a workflow view to see each agent getting called, and tabs for messages, timeline, metrics, and exports. You can also download the final output as PDF, JSON, or Markdown and pass it to another team or another agent. I also walk you through the real setup like I do in production: cloning starter/final code, understanding the folder architecture (agents, runner/streamer, exports, hooks, components), and wiring the backend with Supabase. We create a new Supabase project, grab URL + keys, run the SQL schema, then install dependencies and run the app on localhost. After login/signup, you add your OpenAI/Claude/Gemini key (stored in your browser, not on my server), pick priority to control token usage, and test live tasks like “write a marketing strategy” or “full stack application architecture.” If you want a portfolio project that proves you can build agentic pipelines end‑to‑end, this one is it.