Docs  Computer use & browser

Mori docs

Computer use and the browser

Mori drives your apps by their buttons and names, works in your own Chrome, and uses the screen only when the job truly needs it, with a bar at the top of the screen that says what she is doing.


The screen comes last

When there is a direct way, Mori takes it: a tool, a connected app, a command, or the app's own scripting. She looks at the screen and moves the mouse only when you ask for the screen itself, or when the work is something that only happens on screen.

Buttons by name

Mori reads the buttons, fields and menus of the front window from your system's accessibility tree and presses them by name, with any model and no screenshot. She lists running apps, opens, focuses and closes them, and reads or changes a window's size and place. When the names run out, she falls back to looking.

The vision loop

When she does need to see, Mori takes a screenshot, decides the next step, acts and checks, step by step, on any model that reads pictures, with an OpenRouter or Anthropic key or a plan's credits. On the Mac she asks for Screen Recording when she starts and for Accessibility on her first click. Stop halts the mouse at once.

The bar at the top of the screen

While Mori uses your computer, a slim bar sits at the top centre of the screen, under the menu bar. It carries her logo, a sentence for the step ("Mori is clicking in Finder", "Mori is typing in Notes", "Mori is looking at the screen", or "Mori is using your computer" between steps), a timer and a red Stop. It moves aside so it never covers the spot being clicked, and it stays up for the whole run. The menu bar icon has a Stop too.

Which apps she may press

  • Look only: banking and trading apps, and Mori's own window.
  • Click only: terminals and code editors.
  • Full control: everything else.

No setting changes a tier, and a refused step says why. See Security.

Your own Chrome

With the Mori Chrome extension connected, web work runs in the Chrome you already use, signed in as you. Mori reads the page, acts on it, takes screenshots, manages tabs and reads the console, from the same thread on the desktop.

  • Pairing uses a signed code. Signed in to the same account on both sides, it is shared for you; otherwise you paste it once.
  • If this computer was ever paired, Mori opens Chrome herself and waits for the extension.
  • On Windows, the system's own UI Automation comes first and the extension is the fallback.

Settings > Chrome chooses where web work runs: Let Mori pick, Mori's Chrome (her own signed-in window) or My browser. One message can name a different one for that turn.

The built-in browser

Without the extension, a built-in browser agent reads a page's elements and clicks by selector, with an Anthropic key, or an OpenRouter key on Sonnet 5.5. With neither, the turn carries on in the normal tool loop.

Shopping and the pay button

Mori compares real listings, adds the right item to the cart and fills in shipping up to the final screen.

  • In your Chrome and on your screen, she never presses a final pay or place-order button, and never types card details or passwords. She stops there for you.
  • Her built-in browser stops at the order review and shows Place the order. She presses pay once, only after you tap it or tell her to.

Screen awareness

Settings > Permissions > Screen awareness lets each message tell Mori which app and window you are in, your front Chrome tab and which terminals are open: titles and addresses only, never a picture of your screen and never a page's text. A preview shows exactly what she would see right now, and one switch turns it off.