The browser became the place where work happens. Our applications, conversations, documents, tools, accounts, and increasingly our AI all live inside it.
But the browser itself still thinks in tabs. Every tab is treated as a separate destination. It does not know why the tab is open, how it relates to the others, what you are trying to accomplish, or whether the work ever reached its goal.
So we do it ourselves. We carry context from one tab to another. We repeat the same intent to different models. We supervise disconnected agent sessions. We remember what happened, what remains unfinished, and what must happen next.
The browser gave us access to the world's software. It left us to hold the work together.
We believe a browser can be much more than tabs. It can move from a passive surface where people navigate software to an active surface that understands the work, carries it forward, and protects the intent behind it.
It can become the trusted custodian of the work.
The browser is uniquely positioned for the age of AI. No other tool sits closer to intent while it is forming. The browser sees what you read, which tools you use, where you return, what you leave unfinished, and the sequence of actions behind the work.
And no other tool sits closer to permission. Your accounts, your sessions, your credentials — the standing you have earned across the web over years — already live here. An agent anywhere else is a capable stranger: it can reason about your work, but it cannot act as you. Inside the browser, with your consent, it can.
This is the part that is easy to miss. Capability is the demo. Permission is the product.
That makes the browser more than a container for AI. It can become the computer AI works through.
Agents will be native to this browser, not attached beside it. They will have an environment built for long-running work: parallel execution, persistent projects, secure credentials, explicit permissions, memory, routines, and the freedom to use the right model for each task.
Today, AI begins when someone stops working and writes a prompt. This browser will understand much earlier. It will know why your tabs are open — how the current page relates to what came before, what goal sits behind the activity, what has already happened, and what remains unresolved.
When you return tomorrow, it will not merely restore the windows. It will restore the intent. It will anticipate the next useful page, question, decision, or agent before you have to search, explain, or begin again.
The unit of work is no longer the tab. It is the goal.
The browser of the past waited for instructions. This one will hold responsibility for continuity.
It will remember the goal as the work moves across pages, applications, people, models, and time. It will preserve the decisions already made, coordinate the agents doing the work, keep routine execution moving in the background, and bring the exceptions — the moments where judgment matters — back to you.
People should set intent, exercise taste, make the trade-offs, and own the outcome. They should not spend their best hours copying context, restarting agents, or checking whether every step was completed.
Humans make the calls. The browser carries everything between them.
A browser capable of carrying real work must earn the right to do so. You choose the models, the context, the tools, the accounts, and the permissions available to every task.
An agent in this browser is hired, scoped, approved, and on the record. This is not autonomy at any cost — it is capability governed by trust.
Today's browser remembers where you went. A finished task is not necessarily a successful one, so the work stays connected until the goal is achieved.
And when a path leads to success again and again, the browser keeps it. Not as a static template, but as a proven way to move similar work forward.
The next person, or the next agent, will not begin with an empty prompt. They will begin with what has already worked.
Over time, the browser learns which context matters, which decisions lead to better outcomes, where human judgment changes the result, and which paths can be trusted again.
Traditional autocomplete learns how you write. This browser will learn how your work succeeds.
The first browser helped people navigate the web. The next will help people and AI accomplish goals through it.
We are not adding AI to the browser. We are changing what a browser is.
Built on the most powerful foundation for AI agents.
Designed to understand and anticipate your intent.
Trusted to carry the work forward until the goal is achieved.