agent-browser is an open-source command-line tool from Vercel Labs that lets an AI agent control a web browser, one short command at a time.
In plain words
The agent asks for a snapshot and gets the page back as short text, in which every element on the page gets a tag like @e1. It clicks or types on a tag, takes a fresh snapshot to see what changed, and keeps going. Because the browser keeps running between commands, every step takes only one quick shell call. By default it drives Chrome for Testing over the Chrome DevTools Protocol. It is written in Rust and uses the Apache 2.0 licence.
A quick example
Imagine you ask a coding agent to check that your sign-up form works. It opens the page, takes a snapshot, types an email into the field tagged @e2, clicks the button tagged @e3, then takes another snapshot to confirm the thank-you message appeared.
Where you'll meet it
- Letting coding agents test and use websites from the terminal.
- Filling in forms, taking screenshots and checking pages in automated agent runs.
Related
Sources
- Docsvercel-labs/agent-browser, Vercel Labs (GitHub) · read 7 Nov 2026
- Docsagent-browser, agent-browser docs · read 7 Nov 2026
Checked by WhatIsAI against the sources above · 7 Nov 2026. Spotted an error? Report it.