Skip to main content
There are two fast paths:
  • Use stealth-extract when you only need to read page content.
  • Use a browser session when you need to click, type, log in, or inspect state between steps.

Path A: Extract content

Use stealth-extract for JS-rendered pages, protected content, and one-off collection.

Path B: Automate a browser

Open a browser session, inspect page state, and interact with elements by index.

Path A: extract content

Use this when basic fetchers cannot handle JavaScript rendering, anti-bot checks, or region restrictions:
The command opens a stealth browser, waits for the page to render, returns content as Markdown, and closes the browser. You do not need to manage a browser or session. Common options:
Use this path for protected content, batch collection, and information retrieval. See Anti-detection & Blocking.

Path B: automate a browser

Use this path when the task requires interaction.

The core loop

  1. Open a page with browser open.
  2. Read state to see the current URL, title, and indexed elements.
  3. Interact with click, input, select, and related commands.
  4. Read state again after the page changes.
  5. Repeat until the task is complete.
  6. Close the session.

Understand state

state is the agent’s view of the page. It returns compact text with indexed elements:
Each [N] is an element index:
Indexes can change after navigation, page updates, or DOM changes. Call state again before acting on a changed page.

Chain commands carefully

[!TIP] Chain commands only when each step is independent. If the next action depends on page output, run state first and decide from the result.
Use && for independent steps:
Run commands separately when the next action depends on intermediate output, such as reading state before choosing what to click.

Learn more

Browser Modes

Pick Chrome, chrome-direct, stealth privacy, or stealth fixed identity.

Anti-detection & Blocking

Use stealth, CAPTCHA solving, proxy options, and remote handoff.

Better Headless Browser

Keep browser work headless while preserving stealth and handoff.

Concurrency & Isolation

Separate browsers, sessions, accounts, and one-off jobs.

Command Reference

Open the full CLI command index.

Skill Forge

Turn a successful exploration into a reusable scraping skill.