I recently managed to get my hands on Browserbase. The platform that allows you to start a session with browser and get to extract data, manipulate browser for certain actions and build apps based on that data.
Here's what you need to build a browser AI agent.
- Google Antigravity IDE
- Browserbase account
You can use any agentic coding tools so it's not something you're limited to. Its just my preference. You can use other browser APIs too so don't limit yourself to a provider.
I managed to give the whole process in this video below which you can check out step by step.
First thing is make sure to sign up to the Browserbase platform. And save the API key that you can get from the dashboard.
Second thing you can do is open agentic coding tool. I am using Antigravity IDE as it is easy to use like cursor. And it has Gemini 3.5+ onwards models to work with.
Third thing is you need a prompt. You have to explain your goal to the tool.
"Create a single page app that would scrape the Google search results for certain business. And extract data and displays it in an accessible yet minimal and presentable ways.Use the browserbase API key (attach). And here are the docs to browserbase: ref1, ref2"
From here onwards Geminin would give you an implementation plan. And you also get the walkthrough once the app is generated. Check if it does what it is intended to do. Also this is just one shot implementation. So at the end you have to refine it for final touch.
Above image is the final result of my prompt. And it is pretty decent app which i restricted to a certain business type.
Like this you can create many browser agents that serve a specific purpose.
Top comments (0)