Use agent-browser to start and interact with a real web browser for testing, scraping, or verifying web-based features
I help you use agent-browser to launch a headless Chromium instance and interact with real web pages. agent-browser is a browser automation CLI that uses a snapshot-and-ref workflow optimized for AI agents — you navigate, snapshot the accessibility tree, then interact with elements by ref.
Use this skill when you need to interact with a real web browser — for example to verify a web UI, scrape content, test a local dev server, or automate a multi-step browser workflow.
agent-browser --help to review the available commands.agent-browser open <url> to navigate to a URL (e.g. http://localhost:3000 or any public site). This launches the browser automatically if not already running.agent-browser snapshot -i to get the accessibility tree with interactive elements and their refs (@e1, @e2, etc.).agent-browser click @e1 — click an elementagent-browser fill @e2 "text" — clear and fill an input fieldagent-browser get text @e1 — extract text from an elementagent-browser get attr @e1 href — get an attribute valueagent-browser hover @e3 — hover over an elementagent-browser select @e4 "value" — select a dropdown optionagent-browser snapshot -i again to get updated refs.agent-browser screenshot [file] — take a screenshotagent-browser eval "<js>" — evaluate JavaScriptagent-browser wait "<selector>" — wait for an element to appearagent-browser wait --load networkidle — wait for network to be idleagent-browser get title / agent-browser get url — inspect the current pageagent-browser tab — list open tabsagent-browser tab new <url> — open a new tabagent-browser back / agent-browser forward / agent-browser reload — navigationagent-browser close to shut down the browser.The recommended workflow for AI agents is:
agent-browser open <url> — navigateagent-browser snapshot -i — get interactive elements with refsagent-browser click @e1 / agent-browser fill @e2 "text" — interact using refs@e1, @e2, etc.) are deterministic identifiers from the most recent snapshot. Always re-snapshot after navigating or triggering page changes.agent-browser snapshot -i -c for a compact view of interactive elements only.agent-browser snapshot -i --json for machine-readable output.agent-browser click "#submit".agent-browser is visible "<selector>" to check element visibility (exit code 0 = visible).