Puppeteer MCP Skill

🎯 Goal

Enable the AI Agent to control a headless browser for testing, scraping, and UI verification using the official Model Context Protocol standard.

🛠️ Tools

navigate

Navigate the browser to a specific URL.

• url: The URL to visit (string)

screenshot

Take a screenshot of the current page or a specific element.

• path: Path to save the screenshot (string)
• selector: CSS selector to capture (optional, default: full page)

click

Click an element on the page.

• selector: CSS selector of the element to click (string)

fill

Fill an input field.

• selector: CSS selector of the input (string)
• value: Value to type (string)

evaluate

Execute JavaScript in the page context.

• script: JavaScript code to run (string)

get_content

Get the HTML content of the page.

• selector: CSS selector to get content from (optional, default: body)

🚀 Usage Rules

1. Headless: Creating a session spawns a headless browser by default.
2. Selectors: Use stable CSS selectors (id, data-testid) where possible.
3. Wait: Ensure standard Puppeteer waitFor logic is considered if the page is dynamic (handled by the tool implicitly or explicitly).