How to use the Puppeteer MCP server for web scraping

Open your MCP host's configuration file and add a playwright entry. The server uses the @anthropic-ai/playwright-mcp package, which is the official Anthropic-maintained browser automation server. It uses npx with the -y flag for automatic installation. On first run, Playwright will download the browser binaries it needs (Chromium by default).

1{
2  "mcpServers": {
3    "playwright": {
4      "command": "npx",
5      "args": [
6        "-y",
7        "@anthropic-ai/playwright-mcp"
8      ]
9    }
10  }
11}

Fully quit and reopen Claude Desktop, Cursor, or VS Code. The Playwright server will start and its browser automation tools will become available. You should see tools for navigation, screenshots, clicking, typing, and JavaScript execution in the tools list.

Start with a simple navigation task. Ask the AI to visit a URL and describe what it sees, or take a screenshot. The AI will launch a browser window (visible on your screen), navigate to the page, and capture the rendered result. This confirms the server is working correctly.

1Example prompts:
2
3"Navigate to https://example.com and take a screenshot"
4
5"Go to my app at localhost:3000 and describe the current state of the homepage"
6
7"Visit https://news.ycombinator.com and tell me the top 5 stories right now"

The AI can interact with web page elements by clicking buttons, filling input fields, selecting dropdowns, and submitting forms. Describe the action you want in natural language and the AI will find the right element on the page and interact with it.

1Example prompts:
2
3"Go to localhost:3000/login, fill in the email field with 'test@example.com' and the password with 'password123', then click the Sign In button"
4
5"Navigate to my app, click the 'Create New Project' button, and fill out the form with a sample project name"
6
7"Go to localhost:3000/settings and toggle the dark mode switch"

The AI can extract text content from specific page elements, scrape structured data, and execute custom JavaScript in the browser context. This is useful for testing, debugging, and data extraction. The AI can inspect the DOM, check console errors, and verify page behavior.

1Example prompts:
2
3"Go to localhost:3000 and check if there are any JavaScript errors in the browser console"
4
5"Navigate to localhost:3000/dashboard and extract all the data from the stats cards into a table"
6
7"Visit my app, open the browser dev tools network tab, click the Submit button, and tell me what API calls are made"

For VS Code with GitHub Copilot, use the servers key instead of mcpServers. Create a .vscode/mcp.json file in your workspace root.

1{
2  "servers": {
3    "playwright": {
4      "command": "npx",
5      "args": [
6        "-y",
7        "@anthropic-ai/playwright-mcp"
8      ]
9    }
10  }
11}