Skip to main content
Browser-act gives agents a practical escalation path for sites that block basic fetchers or standard browser automation.

Three-layer strategy

Environment layer

The stealth browser is designed to look like a normal browser session instead of a script.

Dynamic proxy

List available regions:

Custom proxy

Custom proxies support HTTP and SOCKS5. You own the proxy quality, identity, and rotation strategy.

Privacy mode

Use privacy mode when each session should start clean:
[!TIP] Use privacy mode for one-off scraping, avoiding fingerprint buildup, and separating account identities. The trade-off is that login state is not preserved.

Execution layer

stealth-extract

Use stealth-extract when the task is read-only:
It opens a stealth browser, renders the page, returns Markdown or HTML, and closes the browser.
Use stealth-extract for reading. Use a browser session when you need interaction.

solve-captcha

Use solve-captcha when a session reaches a supported CAPTCHA challenge:
A successful result returns solved=True.

Escalation flow

Human layer

When automation cannot move forward, remote-assist creates a live URL. A human can open the link on any device, complete the challenge, and let the agent continue in the same session. See Remote Assist.
  1. Start with chrome or chrome-direct for ordinary logged-in work.
  2. Move to stealth with a proxy when blocking appears.
  3. Use solve-captcha if a supported CAPTCHA appears.
  4. Use remote-assist when human judgment, 2FA, hardware tokens, or unsupported challenges are required.

Compared with basic tools

Learn more

Better Headless Browser

Run silently while keeping stealth and handoff available.

Browser Modes

Choose the right browser identity for the job.

Skill Forge

Turn site-specific knowledge into a reusable skill.