Agent Browser

About

Automate the browser from a CLI built for AI agents.

Agent Browser is a browser automation CLI for AI agents. Snapshots return the accessibility tree with refs, so an agent picks the exact element every time, and compact text output uses far fewer tokens than JSON or the full DOM. It has more than 50 commands for navigation, forms, screenshots, network, storage, files, tabs and frames, with video recording, streaming, debugging, profiling and diffing built in. Sessions, profiles, auth state and cookies support long-running agents.

Features & use cases

• Select elements by ref from an accessibility snapshot
• Save tokens with compact text output
• Run 50+ commands for forms, tabs, network and storage
• Record video and debug sessions
• Keep sessions, profiles and auth state for long runs

Who reaches for it

• Developers building AI agents that need to use a real browser
• Teams testing web apps with an agent and wanting deterministic element selection

Getting it

Agent Browser has native binaries for macOS, Linux and Windows. It is a Rust CLI with a native daemon that drives Chrome through the Chrome DevTools Protocol, and it publishes documentation and has a pricing page.

Similar tools