Browser Agent
Also called Web agent, AI browser agent, Browser automation agent
An AI agent that uses a web browser to complete tasks: opening pages, clicking, filling forms and reading results on your behalf.
What is a browser agent?
A browser agent is an AI agent that works inside a web browser. You describe a task, such as "find three flights under $400 and list their times" or "fill in this supplier form from my notes", and it navigates websites to do it, step by step.
Browser agents appear as features in AI assistants, as AI-first browsers, as browser extensions, and as developer libraries for building your own.
How does it work?
The agent repeats a simple loop: look at the page, decide the next step, act. It "looks" by reading the page's structure (its links, buttons and form fields), by taking screenshots, or both. Then it clicks, types, scrolls or opens a new page, and checks the result before choosing what to do next.
What is it commonly confused with?
- A computer-use agent. That controls a whole computer, including desktop apps. A browser agent is limited to the web, which makes it easier to sandbox.
- AI search. An AI search engine reads pages and summarises them. A browser agent takes actions on pages: submitting forms, adding items to a basket, changing settings.
Why does it matter?
A lot of everyday work happens on websites without an API, so browser agents can save real time on research, data entry and admin. They also bring new risks. Web pages can hide instructions meant for the agent (prompt injection), and an agent signed in to your accounts can do real damage if it's misled. Good tools ask for your approval before purchases, sending messages and other irreversible steps (human in the loop), and you should check that a tool does this before you rely on it.
