We’re introducing Browser Agent, a new way to experience AI that doesn’t just answer questions — it takes actions for you right in your current browser. This feature is now available as a research preview for a limited group of users.
From chat to action
You’ll find Browser Agent in your sidebar. It’s a new widget that lets you describe a task in plain language and then watch as the AI performs it step by step on the webpage. You can stop the agent anytime, keeping full control.
Why we built it
So much of modern work happens in the browser — managing tools, filling forms, checking data, or testing content. Until now, using an AI agent to perform these actions often requires switching to a dedicated “agent browser.” With Browser Agent, you can now try AI automation directly in your existing browser, without changing how or where you work.
What to expect
Browser Agent is still in research preview, so it may be slow or inconsistent as we continue exploring what’s possible and safe. Because it’s computationally expensive, access is currently limited to Ultra users. Insights from this phase will help us improve speed, safety, and reliability before expanding availability.
See it in action
We’ve prepared a short intro video showing Browser Agent in use — from describing a task to watching it complete actions step by step.
Share your feedback
We’d love to hear what works well, what doesn’t, and what you’d like to see next. You can share feedback through our feedback form. Every response helps us make Browser Agent smarter, safer, and more useful for everyone.