← Concept Index

Agents — core

Computer and browser use

Also called: computer use, browser agents, operating the screen

DEFINITION

Giving an agent the ability to operate a computer the way a person would — reading the screen, moving the cursor, clicking, typing, and navigating a browser — so it can use software that has no dedicated tool or API.

WHY IT MATTERS

It widens what an agent can reach enormously, and widens the risk with it: anything visible, including a web page or document, can carry instructions, and the agent acts with the user's full access. It's the strongest case for narrow permissions and approval on consequential actions.

COMMONLY CONFUSED WITH

A defined tool call. A scoped tool does exactly one thing; a computer-using agent can do anything the logged-in user can — a far larger blast radius.