| name | browser |
| description | Interactive browser automation with agent-browser. Open pages, click elements, fill forms, take screenshots. Use for testing UI, automating tasks, or debugging visual issues. |
| allowed-tools | Bash, Read |
| argument-hint | <url> [action] |
Browser Automation
Interactive browser control using agent-browser (Playwright-based).
Quick Reference
pnpm exec agent-browser open <url>
pnpm exec agent-browser screenshot [path]
pnpm exec agent-browser screenshot --full [path]
pnpm exec agent-browser snapshot -i
pnpm exec agent-browser click @e1
pnpm exec agent-browser fill @e2 "text"
pnpm exec agent-browser hover @e3
pnpm exec agent-browser press Enter
pnpm exec agent-browser viewport 375 812
pnpm exec agent-browser viewport 1920 1080
Setup (First Time)
cd frontend
pnpm add -D agent-browser
pnpm exec agent-browser install
Workflow
1. Open a Page
pnpm exec agent-browser open http://localhost:8080
pnpm exec agent-browser open https://example.com
2. See What's Interactive
pnpm exec agent-browser snapshot -i
3. Interact
pnpm exec agent-browser click @e1
pnpm exec agent-browser fill @e2 "Hello World"
pnpm exec agent-browser hover @e3
pnpm exec agent-browser press Enter
pnpm exec agent-browser press Escape
4. Screenshot
pnpm exec agent-browser screenshot
pnpm exec agent-browser screenshot ./my-screenshot.png
pnpm exec agent-browser screenshot --full ./full-page.png
5. Responsive Testing
pnpm exec agent-browser viewport 375 812
pnpm exec agent-browser viewport 768 1024
pnpm exec agent-browser viewport 1920 1080
pnpm exec agent-browser screenshot ./mobile.png
Sessions
Isolate tests in separate browser sessions:
pnpm exec agent-browser --session login open http://localhost:8080/login
pnpm exec agent-browser --session login fill @e1 "user@test.com"
pnpm exec agent-browser --session login click @e2
pnpm exec agent-browser --session dashboard open http://localhost:8080/dashboard
Wait for Elements
pnpm exec agent-browser wait visible "[role='dialog']"
pnpm exec agent-browser screenshot ./modal.png
Common Patterns
Test a Form
pnpm exec agent-browser open http://localhost:8080/form
pnpm exec agent-browser snapshot -i
pnpm exec agent-browser fill @e1 "John"
pnpm exec agent-browser fill @e2 "john@example.com"
pnpm exec agent-browser click @e3
pnpm exec agent-browser snapshot -i
pnpm exec agent-browser screenshot ./form-submitted.png
Test a Modal
pnpm exec agent-browser click @e1
pnpm exec agent-browser wait visible "[role='dialog']"
pnpm exec agent-browser screenshot ./modal-open.png
pnpm exec agent-browser press Escape
pnpm exec agent-browser screenshot ./modal-closed.png
Responsive Screenshots
pnpm exec agent-browser viewport 1920 1080
pnpm exec agent-browser screenshot ./desktop.png
pnpm exec agent-browser viewport 768 1024
pnpm exec agent-browser screenshot ./tablet.png
pnpm exec agent-browser viewport 375 812
pnpm exec agent-browser screenshot ./mobile.png
Troubleshooting
Browser not installed
pnpm exec agent-browser install
Page not loading
Check if dev server is running:
lsof -i:8080
Element not found
Use snapshot -i to see current elements:
pnpm exec agent-browser snapshot -i
Reference