browser-act/skills is a browser-automation resource that gives AI agents control of real browsers with stealth features, CAPTCHA solving, human handoff, and isolated concurrent sessions. It is designed for LLM reasoning rather than human-written scripts.
Project overview
It addresses three persistent gaps in AI agent browsing—anti-bot wall bypass, human handoff for stuck tasks, and cross-contamination-free parallel sessions—while delivering indexed text output that is several times more token-efficient than JSON or HTML.
Project type
AI Agent · AI Coding · Data Processing
Use cases
Coding & Development · Automation
Deployment
Refer to project documentation
License
MIT
Best for
AI agents that need real-browser control with stealth features and isolated concurrent sessions.
LLM-based workflows optimized for reasoning over token-efficient indexed text rather than raw HTML or JSON.
Key capabilities
Provides indexed text output and CLI commands for agents to navigate and interact with web pages without DOM parsing.
Automatically solves CAPTCHAs via the solve-captcha command.
Explores a website once, discovers APIs and data patterns, and generates a deploy-ready scraping Skill package.
Pulls protected pages in a single command without configuration.
Limitations and risks
Advanced stealth features and dynamic/static proxies require payment. These are optional, managed, paid external services.
CAPTCHA solving and anti-bot wall circumvention features carry a potential for misuse; adopters should evaluate appropriate-use policies before deployment.
Getting started
Tell an AI agent: Install browser-act. Use the skill source at https://github.com/browser-act/skills/tree/main/browser-act. No manual coding is required.