An AI agent can now control a web browser directly from the command line. This tool is built in Rust for speed and uses the Chrome DevTools Protocol. It runs without Node.js or Playwright.
The headless browser can navigate to any page. It can click, fill forms, and take screenshots. A special accessibility snapshot gives each element a unique @ref label. This makes it easy for an AI to interact with the page.
You can also scrape web data and batch commands. This skill is made for AI agents that need to automate web tasks.
Global
mkdir -p ~/.claude/skills/agent-browser-automationProject
mkdir -p .claude/skills/agent-browser-automationSource Repository
Agent Browservercel-labs/agent-browser
Fast browser automation tool for AI agents to control any website
Develop Userscriptsxixu-me/skills
Build and debug browser userscripts for Tampermonkey and ScriptCat
Just Scrapescrapegraphai/just-scrape
Search, scrape, crawl, and monitor web pages with an easy command line tool
Browser Actbrowser-act/skills
AI agents use browser-act to automate web tasks like fetch forms screenshots and complex workflows
Firecrawlfirecrawl/cli
Search, scrape, and crawl websites with one simple command line tool
Playwright Climicrosoft/playwright-cli
Automate browser tasks and test web pages with simple terminal commands
Browser Usebrowser-use/browser-use
Automate web testing, form filling, and screenshots with simple CLI commands
Browser Act Skill Forgebrowser-act/skills
Explore any website once and create reusable skill packages forever