Give an AI assistant control of a real browser to navigate pages, click, and take screenshots.
Preview full content
Lets an AI assistant drive an actual browser — navigate to pages, click elements, fill forms, take screenshots — useful for testing a web flow, scraping content that needs JavaScript to render, or having the assistant show you what a page actually looks like.
```json
{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-puppeteer"]
}
}
}
```
This is one of the more powerful/riskier servers to grant — a browser it controls can submit forms, click buttons, and navigate anywhere, so only enable it in a session where you're comfortable with an assistant taking real actions on pages, and supervise it on anything involving a logged-in account or a destructive action (submitting, deleting, purchasing).
Check current docs for headless-vs-visible browser options and any sandboxing settings before using it against sites you care about.