Cloudflare Playwright

You can run a Model Control Protocol (MCP) server that uses Playwright to automate a browser through Cloudflare Workers. This lets you issue high-level actions (navigate, type, click, take screenshots) to control a browser, enabling AI assistants to perform web automation tasks across supported platforms.

How to use

To use this MCP server with an MCP client, connect to the MCP URL and start issuing actions that map to browser operations. You can navigate pages, fill forms, click elements, take screenshots, and more. Start with simple tasks to verify the flow, then compose longer automation sequences that combine multiple browser actions.

How to install

Prerequisites you need on your machine are Node.js and npm. You also need the Cloudflare Wrangler tool to deploy the worker. Ensure you have access to the internet from the terminal where you run commands.

# Install dependencies locally
npm ci

# Deploy the MCP server to Cloudflare Workers
npx wrangler deploy

# Optional: verify deployment and note the produced URL (e.g., https://your-workspace.your-worker.workers.dev/sse) for MCP clients

Additional notes

The server is designed to integrate with MCP clients such as Cloudflare AI Playground or Claude Desktop via local proxies. For local testing with Claude Desktop, you can proxy the remote MCP server using mcp-remote, which exposes it locally. The two main ways to connect are via HTTP to the public MCP URL or via a local stdio configuration that runs a proxy command.

Available tools

browser_navigate

Navigate the active browser to a specified URL and update the page snapshot to reflect the new location.

browser_type

Type text into a focused input field or element within the page.

browser_clickt

Click a target element on the page, such as a button or link.

browser_take_screenshot

Capture a screenshot of the current browser view and return it in PNG format.