Playwright MCP Browser Automation

Core Workflow

code

navigate → snapshot → interact → snapshot → close

Tool	Use Case
`playwright_browser_navigate`	Open URL
`playwright_browser_snapshot`	Read page structure (preferred over screenshot)
`playwright_browser_take_screenshot`	Capture visual state
`playwright_browser_click`	Click elements
`playwright_browser_type`	Type text into elements
`playwright_browser_fill_form`	Batch fill multiple form fields
`playwright_browser_select_option`	Select dropdown options
`playwright_browser_press_key`	Keyboard shortcuts
`playwright_browser_drag`	Drag and drop
`playwright_browser_hover`	Mouse hover
`playwright_browser_evaluate`	Execute JavaScript
`playwright_browser_file_upload`	Upload files
`playwright_browser_handle_dialog`	Handle alerts/confirms
`playwright_browser_wait_for`	Wait for text/disappear/timer
`playwright_browser_tabs`	Tab management (list/new/close/select)
`playwright_browser_console_messages`	Check for errors
`playwright_browser_network_requests`	Inspect network traffic
`playwright_browser_run_code`	Run Playwright code snippets

Order matters:

Form filling: Use fill_form with fields array (each needs name, type, ref, value).

Element references: Always use ref from snapshot output - never guess selectors.

Dialogs: Call handle_dialog with accept: true/false before triggering action.

Errors: Check console_messages(onlyErrors=true) before completing.

Screenshots: Only use when visual inspection needed. Use snapshot for structure.

Multiple tabs: Use tabs(action="new/select/close") with index for close/select.