Term
计算机操作能力
Next Token / Wiki

Computer Use

Computer use is the ability of a model to operate a computer's graphical interface the way a person does — reading the screen, moving the cursor, clicking and typing — to complete tasks across applications. Anthropic and OpenAI both ship computer-use tools or model capabilities.

Also known as

computer use agent · CUA

Definition

Computer use is the ability of a model to operate a computer’s graphical interface the way a person does — reading the screen, moving the cursor, clicking and typing — to complete tasks across applications. Anthropic and OpenAI both ship computer-use tools or model capabilities.

How it works

It runs as a loop: capture a screenshot, have the model decide on actions such as clicks and keystrokes, execute them on an underlying virtual machine or container, then feed the next screenshot back until the task is done.

Usage

Legacy software without APIs, repetitive work spanning several applications, and end-to-end interface testing are common assignments for computer-use agents.

Compared with browser use, the scope is broader: computer use addresses the whole desktop, while browser use stays inside the browser. It can also be seen as the digital counterpart of embodied AI.

Sources