BrowserAct Skills is a browser-automation command-line tool that AI agents load as a Skill to drive real browsers for automation, web data extraction, and account-based workflows. It works with any agent that can run shell commands and load Skills, on Windows, macOS, and Linux. BrowserAct also offers a cloud mode in which you describe the site and fields you need, and it builds, tests, and runs a reusable scraping bot in a cloud browser.
Blocking is handled in three layers: an environment layer (stealth fingerprints, TLS rotation, proxy switching), an execution layer (automatic CAPTCHA solving and a one-command extractor for protected pages), and a human layer, where a remote-assist link lets a person take over the browser from any device and hand control back to the agent.
Three browser modes cover different scenarios:
| Mode | Scenario | Key trait |
|---|---|---|
| chrome | Reuse local Chrome login state | Profile import or CDP attach |
| stealth privacy mode | Batch scraping without login | Fresh fingerprint per session and proxy rotation, no residue |
| stealth fixed identity | Logged-in accounts, multi-browser parallel | Stable fingerprint and IP for a stable account identity |
The interface is designed for LLM reasoning rather than human-written scripts. Page state comes back as a compact indexed text list, so an agent clicks or types by index without parsing the DOM. Each browser carries a description matched to tasks by meaning, and named sessions keep parallel agents from interfering. Sensitive operations such as creating or deleting browsers, importing profiles, and changing proxies require explicit user approval each time.
A companion Skill Forge explores a website once, discovers its APIs and data patterns, and generates a reusable Skill package for repeated extraction. A catalog of over 30 pre-built Skills covers sites such as Amazon, Google Maps, YouTube, Reddit, WeChat, and Zhihu. Managed proxies and more than five stealth browsers are paid features.
Features
- Anti-blocking layers: stealth fingerprints, TLS rotation, residential proxies, and automatic CAPTCHA solving
- Remote handoff: a live link lets a person take over the browser from any device, then the agent continues
- Browser modes: reuse local Chrome login state, run fresh privacy-mode sessions, or keep a fixed stealth identity per account
- Isolated concurrency: separate cookies, fingerprints, and proxies per browser, or multiple sessions sharing one login
- Indexed interaction: compact indexed text state so agents click and type by index without DOM parsing
- Confirmation gating: sensitive browser, profile, proxy, and privacy operations need explicit approval each time
- Skill Forge: explores a site once and generates a reusable extraction Skill
- Solutions catalog: over 30 pre-built Skills for sites including Amazon, Google Maps, YouTube, and Reddit
- Cloud bots: describe the data and get a tested scraping bot run in a cloud browser, with output to CSV, JSON, APIs, webhooks, Make, n8n, or Zapier
- Agent compatibility: Claude Code, Cursor, VS Code, OpenCode, OpenClaw, Codex, and Gemini CLI
