Launching today
Jev Ultrafast Agent
AI browser agent without screenshot-heavy automation
0 followers
AI browser agent without screenshot-heavy automation
0 followers
Jev Ultrafast Agent is a Chrome-based AI browser agent that replaces screenshot-heavy automation with an indexed action space. Instead of repeatedly sending screenshots to a vision model, Jev turns interactive page elements into structured inputs and lets the AI execute typed actions like CLICK, TYPE_TEXT, SELECT and SCROLL. It runs directly in Chrome with no Python or local server required. Bring your own model and API key - including Claude and OpenAI-compatible APIs.
Free
Launch Team
ClarifyThe full-stack AI CRM for modern GTM teams
Promoted
Hey Product Hunt 👋
We built Jev because we kept coming back to one question: why should an AI browser agent need to take and process a screenshot every time it wants to click a button?
Jev takes a different approach. It converts interactive elements on a webpage into an indexed action space. The model sees a structured representation of the page, chooses one typed action - CLICK, TYPE_TEXT, SELECT, SCROLL, WAIT, etc. - and the browser executes it.
Then it observes the updated page and repeats.
The goal is simple: make browser agents faster, more deterministic and more token-efficient, while using vision only when it’s actually necessary.
We also wanted to make experimentation easy, so Jev runs entirely as a Chrome extension. No Python environment or local server is required. You connect your own model and API key and start using it directly in the browser.
This is still an early release and we’d genuinely love feedback from people building with AI agents and browser automation.
In particular, we’re curious: where do you think structured browser interaction ends and vision becomes necessary?
Thanks for checking out Jev! 🚀