We turned Jev into a general browser skill for agents: jev-browser. Give it a website and a task. The browser opens automatically, and Jev decides every click based on what's on the screen. Here's a demo:

did you open source this project? would love to feature it in

Thanks! Once we've polished it up a bit more, we'll open source it.

Jev picking every click is wild.

been trying it the last couple days. automating some stuff with it and it’s actually so useful

I was definitely looking for this, where can I find it?

We'll be sharing it inside our community. If you're interested, feel free to join us.

Sure! Will it be an open repo or will be behind paywall?

Do you have a public repo for this? Or is this a closed solution that you built?

We've already shared it in our community. Once it's a bit more polished, we'll consider open-sourcing it too.

seeing the page can choose a click. reliable work still needs a second layer: permission boundaries and a check that the intended result actually happened.seeing the page can choose a click. reliable work still needs a second layer: permission boundaries and a check that the intended result actually happened.

How does it handle unexpected popups or captchas?

the useful boundary is screen state to next action. that makes the same skill portable across sites without hardcoding each flow

A dedicated browser decision layer could make agent automation feel genuinely autonomous.

Where can I get them?

I’ve shared it in our community for now. Planning to open source it later, so keep an eye out.

把浏览器操作变成可评估的决策层，关键是失败回放和可观测性，速度只是结果之一。

I think a lot of automation tasks that involve decision-making could be handled by Jev. It's a very general-purpose decision layer.

The useful leap here is making the browser handle the setup and the clicks, not just generate instructions. Reliability is what will make this kind of agent genuinely useful.

The cost and speed are probably the biggest wins for me.

The important bit is re-reading the page after every click. Otherwise Jev is choosing from a screen that no longer exists.

ещё одно фейковое решение на jev

Could you please share the repo link?

It’s currently shared in our community. I’ll open source it later too, so stay tuned.

Do you have to keep prompting it per task? And does it open a new browser instance each time?

no, you just describe the goal clearly. pretty simple, no complicated prompt needed

이런 것은 무조건 시도해봐야죠

Yeah, it can handle all kinds of general-purpose tasks.