Browser and Desktop
Use a web browser and a full desktop on the agent's Machine, and work in them alongside the agent.
Browser is a web browser on the agent’s Machine that you and the agent can both use, and it stays signed in to the sites you sign in to there. Desktop is the Machine’s whole screen, where any program with a window appears, so you can watch the agent use an app or use one yourself. Reach for them when a Preview isn’t enough: a site that needs your sign-in, an address only the Machine can reach, or an app that isn’t a website.
Open Browser or Desktop
Select Browser or Desktop in the Session’s right sidebar; on a computer they’re under + in the tab row. The first time, it shows Starting Browser… or Starting Desktop… for a few seconds, then opens. Click inside Browser and use it as you would on your own computer. To give it the whole window, select Full screen Browser or Full screen Desktop; press Esc to come back.
Websites in Browser are laid out for the panel, as in a browser window that size, and text stays sharp. Resize the sidebar or go full screen and the page reflows to fit. A panel narrower than 500 pixels shows a page 500 pixels wide, shrunk to fit.
Use Desktop
Desktop opens view-only, so you can watch without getting in the agent’s way. Its toolbar says whether it’s Live or Reconnecting….
- Interact: hover over Desktop and select Interact to use it with your mouse and keyboard. On a touch screen, tap it first. Interact with Desktop in Search does the same. Select Done, or click anywhere else in Polaris, to go back to watching. In full screen, Desktop takes your clicks and typing without Interact.
- Desktop size: Fit to panel, the default, shows the whole screen. Actual size makes text readable; drag or scroll to move around. Device pixels is sharpest on high-resolution screens.
- Sound: Desktop plays the Machine’s sound, but starts muted each time you open it. Select Unmute audio, then use Desktop volume. If there’s no Unmute audio, that Machine has no Desktop sound; a new Machine has it.
While Browser or Desktop has your keyboard, your keys go to the Machine and Polaris shortcuts pause. Click outside it, such as in the composer, to use them again.
Float Browser or Desktop
Float keeps Browser or Desktop in a small window above your other apps and tabs, so you can watch the agent while you work somewhere else. It needs a Chromium-based browser, such as Chrome or Edge; other browsers don’t show Float.
Select Float Browser or Float Desktop in its toolbar or in the Session palette. The window shows the same view, and the sidebar says it’s in a floating window.
- Use it: click inside Browser as usual. Desktop stays view-only until you select Interact in the window’s header; select Done to go back to watching.
- Bring it back: select Back to the panel in the window, or Bring back Browser or Bring back Desktop in the sidebar or the Session palette. Close closes the window, and the view goes back to its tab in the sidebar.
- One at a time: floating the other view moves it into the same window.
Moving the view between the sidebar and the window reconnects it for a moment. The window closes when you leave the Session.
Choose the right view
- Preview: click through the app the agent is running, quickly, beside the Session or in a tab of your own browser.
- Browser: open any website as the Machine sees it, including your dev server and sites you sign in to. In a Preview tab’s menu, Open in Browser moves the page there. You can also open a Session link in the side panel.
- Desktop: use or watch a program with a window, such as a browser the agent opens to test your app.
Browser and the browsers on Desktop are separate, each with its own sign-ins.
Sign in once for the agent
Browser keeps cookies and sign-ins on the Machine, so they’re still there after the Machine sleeps and wakes, though a site can still ask you to sign in again. Sign in to a site in Browser yourself, and the agent can work there in the same browser without your password. When Browser starts again, it opens the last website it showed.
To keep agents out of Browser on all your Machines, turn off Agents can use Browser in Settings → General. Turn it back on to let them use it again.
Browser plays websites’ sound in your browser tab. Use Mute audio or the Browser volume slider in its toolbar; they change only what you hear.
Watch the agent and take over
When the agent uses a website, it works in Browser, so you see the pages it opens. While it acts, Browser shows Agent is browsing under its toolbar and a labeled pointer where it clicks. A tab the agent opens shows in Browser right away. Your own clicks always go through at once.
To use Browser without the agent acting at the same time, for example to sign in or fill in a form, select Take control next to Agent is browsing. Browser then says You have control. The agent can still see the pages, but it can’t click, type or open pages until you select Hand back. If it tries, Browser says The agent is waiting to use Browser. Closing Browser doesn’t hand control back. Take control of Browser and Hand back Browser in the Session palette do the same. If there’s no Take control, that Machine can’t hand you control; a new Machine can.
When the agent runs a program with a window, it appears on the Desktop, and the agent can use it there: it reads the app’s windows, clicks and types. Open Desktop to watch it work. The agent’s own pointer shows where it acts, and yours stays where you left it. Some Machines can’t let the agent use Desktop apps; a new Machine can. You and the agent share the same screen, and nothing locks either of you out, so if you both type at once, the app gets both. Take over when the agent asks you to, for example to sign in or solve a CAPTCHA. To take over sooner, stop the agent first. Apps on the Desktop can close when the Machine sleeps, so save your work in them before you leave.
To paste text from your computer, click inside Browser, or in Desktop after Interact, where the text should go, and press Ctrl+V, on a Mac too. Allow clipboard access if your browser asks. Only text pastes. Browser copies text only into the Machine; to bring text from Desktop to your computer, see below.
Copy from Desktop to your computer
- Select Interact on Desktop, or open it full screen.
- Copy text in a Desktop app, for example with Ctrl+C.
- Copied on the Desktop appears over Desktop with the text you copied. Select Copy to your computer.
Only the text shown there is copied, and only when you select it. Characters you couldn’t otherwise see or tell apart, such as a line break at the end, invisible Unicode characters, or a Cyrillic letter in a Latin word, are marked in the text so you can check them first. The agent’s copies never reach your clipboard on their own, and text you copy while only watching isn’t offered. Text longer than about 256 KB is too long to copy this way. Some Machines can’t copy to your computer: if Copied on the Desktop never appears, save the text to a file on the Machine and open it from Files instead.
If it doesn’t work
- Browser is still starting or Desktop is still starting: select Check again.
- Browser didn’t start or Desktop didn’t start: select Try again.
- Reconnect Browser, or Desktop disconnected with Reconnect: the connection dropped. Select it to connect again.
- Machine is asleep: select Wake machine.