| name | playwright-mcp |
| description | Use when working with Playwright MCP specifically: browser navigation, accessibility snapshots, element clicks/forms, screenshots, console logs, network requests, and UI regression checks. |
Playwright MCP
Use Playwright MCP for browser-observable behavior.
Flow
- Navigate to the target URL first.
- Capture
browser_snapshot before clicking or filling; use exact element refs from the snapshot.
- Use
browser_wait_for for expected text/state changes.
- Use console messages and network requests after reproducing errors.
- Use screenshots for visual evidence, not as the primary way to locate elements.
Interaction rules
- Prefer accessible roles/names from snapshots.
- Do not use unsafe code execution unless no safe tool can accomplish the task.
- Do not enter real credentials unless user provided test credentials for this task.
- Do not perform destructive UI actions without approval.
Local development
- Check package scripts before starting any app process.
- Do not launch long-running servers with normal bash unless the user requested it and the workflow is safe.
- If the user provides a localhost URL, use it directly.
Validation targets
- Confirm expected route/page renders.
- Check critical interactions, form validation, route transitions, error toasts, and loading states.
- Inspect console errors and failed API calls for UI bugs.
- Test responsive behavior by resizing when layout changes matter.
Cleanup
- Close tabs/pages when finished if no longer needed.
- Report route tested, actions performed, and any console/network failures.
