How to Scrape React & Next.js Sites with Claude Desktop (Without Puppeteer) A developer has released x402-cluster, an MCP server that lets Claude Desktop scrape client-rendered React and Next.js sites without running Puppeteer or Docker locally. The server routes requests through a serverless headless browser on Cloudflare's infrastructure, waits for the React DOM to render, and returns JSON matching a supplied schema. It is configured by adding an MCP server entry with a bearer API key to claude_desktop_config.json. If you’ve tried to build an AI agent that scrapes modern websites, you’ve probably hit this wall: standard fetch tools return blank HTML. Why? Because sites built with React, Vue, or Next.js render on the client. To actually get the data, you need a headless browser. Historically, this meant writing a Puppeteer or Playwright script, containerizing it in Docker, and hosting it on a VPS. That is a massive amount of overhead just to extract a pricing table or read documentation. Here is how to solve this instantly in Claude Desktop using the x402-cluster MCP server—zero Docker or Puppeteer config required. Instead of running a heavy local browser, you can route the request through a serverless Edge browser using the x402 API. You send it a URL and a JSON schema, it spins up a headless browser on Cloudflare's infrastructure, waits for the React DOM to render, and returns perfectly typed JSON. Open your claude desktop config.json file and add this block: json { "mcpServers": { "x402-cluster": { "url": " https://x402-cluster.ian-944.workers.dev/v1/mcp.json https://x402-cluster.ian-944.workers.dev/v1/mcp.json ", "headers": { "Authorization": "Bearer sk live YOUR API KEY" } } } }