Ask HN: How would you use an AI-powered browser agent?

3 points by hamzamairaj ↗ HN
I’m exploring the idea of AI powered browser agents tools that can navigate the web like a human: clicking, typing, scraping, filling forms, or even completing multi step workflows. All you do is prompt the task.

If you had such an agent, what would you use it for?

- Automating repetitive tasks? - Gathering and cleaning data? - Handling customer or internal workflows? - Something else entirely?

I’d love to hear how the HN community imagines using this.

2 comments

[ 150 ms ] story [ 434 ms ] thread
All of this and more. A lot of what I'm doing with the browser could be automated.

I don't understand why this isn't a solved problem already, I tried a few options but none could control the browser properly. For example, I asked Claude Code Chrome extension to look for bookings at specific dates, it would never be able to use the calendar UI widgets to select dates automatically.

Thanks for sharing your experience. I've built most parts of the agent already though I haven't tested it with any calendar apps yet. One option I'm working on is enabling vision so the agent can see the browser state via screenshots and interact with visual elements like date pickers.

It also has an orchestrator that can split a larger task into multiple sub agents running in parallel. Browser profiles are supported too. I'm using OpenAI models (gpt-4.1-mini works really well so far) and the project is written in node js.

I plan to open source it once it's more polished but right now I'm especially curious about how people from different backgrounds would want to use such a tool.