Visit Apps Ved →

browser-act

Browser automation CLI built for AI agents

TLDR

Extract content from a protected page (stealth)
$ browser-act stealth-extract [https://example.com]
Open a page in a session
$ browser-act --session [my-task] browser open [id] [https://example.com]
Read indexed page state
$ browser-act --session [my-task] state
Click by element index
$ browser-act --session [my-task] click [3]
Type into a field by index
$ browser-act --session [my-task] input [2] "[text]"
Load agent skills snapshot
$ browser-act get-skills core --skill-version [2.0.2]

SYNOPSIS

browser-act [--session name] command [args...]

DESCRIPTION

browser-act is a browser automation command-line tool designed for AI coding agents. It emphasizes anti-bot resilience (fingerprints, proxies, CAPTCHA helpers), multi-session concurrency without cookie cross-talk, and compact indexed text output so models can click `3` instead of parsing HTML.Modes include reusing local Chrome login state, stealth privacy sessions (fresh fingerprint), and fixed-identity stealth for multi-account workflows. Sensitive operations can require confirmation gating at the skill layer.

PARAMETERS

--session name

Named session for multi-task isolation and concurrency
stealth-extract url
One-shot fetch of protected page content with stealth mode
browser open id url
Open a URL in a browser instance
state
Return indexed interactive elements (token-efficient for LLMs)
click index
Click the element at the given index from `state`
input index text
Type into the element at the given index
solve-captcha
Auto-solve CAPTCHA challenges when available
remote-assist
Generate a live URL for human takeover, then resume the agent
get-skills
Return environment state, browser list, and available commands for agents

CAVEATS

Full install is agent-oriented (skill packages from the BrowserAct skills repo). Some stealth/proxy features need a free login or paid plan. Automating third-party sites may violate their terms of service—use responsibly.

HISTORY

Open-source BrowserAct Skills project providing a browser layer for agentic tooling on Windows, macOS, and Linux.

SEE ALSO

playwright(1), chromium(1)

RESOURCES

Source code · Homepage · Documentation