agent-browser

Use agent-browser for ordinary browser automation where the agent can own an isolated Chrome session:

• open a URL, inspect the page, click/fill/type/select/upload/scroll
• take screenshots, PDFs, traces, videos, console logs, and page errors
• extract visible text, HTML, attributes, URLs, counts, boxes, styles, cookies, storage, and network requests
• test local web apps with repeatable, command-line browser steps
• run exploratory QA or bug hunts
• automate Electron or Slack when the specialized agent-browser skills get ... content says to do so

Runtime

Installed CLI: