Browser
A browser card is a page embedded in the canvas, not a separate window — and agents only open one with your approval.
The browser card runs a real page inside the canvas. It is an offscreen
WebContentsView, whose frames are painted into a DOM element — that is why
the card respects the board’s stacking and can be captured by a snapshot,
like any other.
Navigating
The address bar takes a URL and navigates on Enter. Back and forward use
the browser’s own history; Alt+←/→ also work, but they are built-in
Chromium navigation, not a configurable Stellar shortcut.
What an agent can do here
A running agent can ask to open a URL, read the page text, click, type, scroll or extract the content. Opening a new URL goes through your consent — in the default flow, each request has a declared reason and a decision from you. Reading a page is passive: it does not ask.
Capture
A browser card’s snapshot is captured at the page’s own resolution, regardless of the board zoom and of whether the card is visible on screen. That is what makes the card useful as evidence in an agent report.