The Hook
“Hey, I am Mati Solina and I want to introduce you Busy. Busy is a place where your AI agents work as a team. Every bot has a role, it gets a private computer in a sandbox, and it's come with instructions and skills. You basically talk like a coworker in Slack, it's the same. Busy is open source and runs on your Vercel account, so the bots are yours and they follow your own rules. If you know WarpBot, Duet, or Dots, the idea is similar. With Busy, the app is yours to change. Need a new feature? Just ship it. Need to connect to your old ERP? Ship that too. You own the code. But let's see how Busy works.”
Product Reveal
“I have here my teams of AI agents divided by departments. Every agent, as I mentioned before, has their own computer, instructions, and context. First of all, you need to connect in Busy your AI models. You can come here to settings, you can use your ChatGPT subscription that you already have, or Groq, you can connect directly, and also you can use AI Gateway. Or if you have the new feature Token Pass from Vercel, this is amazing, you pay just a small fee and get like a big use of very popular models, you can use it too. And here in MCPs, we have several MCPs that you can connect to start working with your agents directly with your account.”
Feature Teaser
“And to spin up a new bot is very simple. You come here to create a new one, you can start from scratch or you can use our templates. The bots has different features. For example, I have like my CTO asking to review a GitHub repository. This is going to run on the on the cloud, so I can just ask for a few modifications and close my Mac, and the agent is going to take care of this.”
Feature Teaser
“We have a video editor built in in the application where you can just upload a video here in Busy, and you can prompt to ask for edits, a new effects, and all this kind of stuff.”
Feature Teaser
“I have a hiring coordinator that has logged in my LinkedIn account, and I'm going to show you how the private computer that each bot has. So here you can see my bot has my LinkedIn account logged in, and the bot is looking for CTO for my for my startup right now. So they have control of their own computer to make this task.”
Call to Action
“And also the bots are asking for permissions and waiting for you to to response to finish their task. And that's it. In next week, we are going to release several changes to add more features. So stay tuned, and let me know if you have any questions. Thank you. Enjoy.”
Synthesizable in Remotion · key components of the developer-screencast-walkthrough template.
Circular-corner rounded rectangle placed at bottom right with persistent speaker feed.
Wrap video stream in AbsoluteFill with bottom and right inline styling, border-radius 24px, and slight box-shadow.
Inner container showing remote OS browser screen loading state transitioning to desktop view.
Use Remotion interpolate on opacity and scale when switching between loading graphic and live browser view.
Purpose. Explain the self-hosted AI agent architecture and ownership model.
Execution. Landing page showing terminal install command and live agent status list.
Purpose. Demonstrate practical multi-agent execution across code, video, and web automation.
Execution. Clicking through agent tabs: GitHub bug checking, video editing, and sandboxed LinkedIn scraping.
Purpose. Reassure users about human-in-the-loop approvals.
Execution. Release Captain prompt demanding user selection before proceeding.