Welcome to Meet.AI, a video calling app powered entirely by AI agents. These aren't just summaries or transcripts. Every meeting here was an actual, real-time call with an AI, trained for a specific role. Whether it's a language tutor, interview coach, sales assistant, or something completely custom, the AI is there to actively help during the call. To start a new call with an AI agent, I Click on the New Meeting button.
I'm going to name this one Startup Coaching Call. I'll also create a new agent here called Hustle Coach. And just for fun, in the instructions, I'll tell it to be direct, hype-driven and a bit over the top, like a founder who just raised a seed round. Once the agent's created, I can select it and I'm ready to start the meeting. With everything set, I click Create Meeting and land on the new meeting page.
Right now, the meeting is scheduled as Upcoming, which means I can come back to it later, cancel it, or start a call whenever I'm ready. So I'll go ahead and hit start. That takes me into a familiar-looking lobby, just like Zoom or Google Meet, where I can check my camera and microphone before joining. Once I enter the call, our new agent, Hustle Coach, shows up right after. This is a live AI, in the meeting with me, responding in real-time, powered by Stream Video SDK and OpenAI Real-Time API.
A hustle coach. What makes a good startup idea? Good startup idea. Alright, listen up. A killer startup idea boils down to 3 things.
Massive market potential, a unique value proposition, and scalability. You want a problem that's begging for disruption, and a solution that screens innovation. And don't forget, your model needs to scale like wildfire. If you can pitch it in one sentence and it sounds epic, you're on to something hot. Now, go out there and disrupt, disrupt, disrupt.
When the call ends, the meeting switches to processing status. Behind the scenes, background jobs kick off automatically, powered by Ingest and Agent Kit. These workflows handle everything. Fetching the transcript, summarizing with AI, and saving the results back to the database. When the meeting is done processing, it's marked as completed, and all the content is ready.
First, there's the summary. A clean breakdown of what was discussed, organized by topic and timestamp. Next is the transcript. It's fully searchable, and when you search for a word, it highlights every place it appears. The whole call is also recorded and available to replay.
And finally, there's a chat-gpt-like interface that actually understands the meeting, powered by StreamChat SDK. You can ask for specifics, like who asked for advice, and it will give you the answer. Meet.ai is a SaaS product, so it runs on a subscription model. Down here, you can see I've hit the limits of the free trial. When I try to create a new meeting, I'm redirected to a page with three upgrade options.
I'll go ahead and upgrade, and that takes me through a quick checkout powered by Pollr. Authentication is powered by BetterAuth, a fully free and open-source auth solution built for modern web apps. It handles email login, social providers, session management, and even a seamless integration with Polar, our payment infrastructure. Our app is also fully responsive. Every screen adapts cleanly to mobile.
Models and drop-downs turn into mobile drawers for a smoother experience on smaller devices. In this course, we won't just write the code. We'll also follow a proper Git workflow throughout the tutorial. Each chapter ends with a new branch and a pull request, reviewed automatically using CodeRabbit AI. Let's quickly go over the test stack we'll be using.
Next.js 15 with React 19, together make our full stack framework supporting server components and server-side rendering. TRPC for full-stack type safety used together with 10-stack query. Drizzle ORM with Postgres provided by Neon for the database. Tailwind v4 for styling, combined with ChatCNUI for accessible and reusable components. BetterAuth for authentication, Polar for payments, Stream for video and chat SDK, Ingest for background jobs, CodeRabbit for AI-powered code reviews, and finally OpenAI for real-time agents and AI-driven features.
And of course, we'll deploy everything to production when it's ready. And now, without further ado, let's get started.