Summary of an article published on The Web Scraping Club on 2026-11-15 (free to read). Read the full article

Every so often someone dismisses chrome-devtools-mcp as a copy of Playwright MCP. I get the reaction. Both are MCP servers, both let AI agents drive a browser, and both show up in the same setup guides. But the differences become obvious once you use them on real tasks, and they matter when you pick a tool for a scraping or automation workflow.

We compare the two side by side. Playwright MCP is Microsoft’s server built on Playwright, centered on browser control and automated workflows. chrome-devtools-mcp is Google’s server built on CDP and Puppeteer, which adds what you normally get from opening Chrome DevTools yourself. We cover how each works, the use cases where each fits, and whether running both makes sense. We also look at how to connect them to the same browser session so state carries over, and where that setup gets harder, for example with remote browsers.

What’s inside

  • Which browsers each server supports, and why chrome-devtools-mcp is tied to Chrome-specific technologies
  • How Playwright MCP exposes the page to the agent through accessibility snapshots instead of screenshots, and how its opt-in tool groups limit tool overhead
  • What chrome-devtools-mcp offers that Playwright MCP does not: performance traces, throttling and emulation, source-mapped console errors, Lighthouse audits, heap snapshots
  • Why installing both servers with separate browsers is the wrong default, and the caveat about concurrent control of one CDP endpoint

In the full article: The comparison tables of tools and coverage, the sample accessibility snapshot output, the tool names and flags for each use case, the shared-browser setup notes, and the FAQ on CLIs and remote browser connections.

Read the full article on The Web Scraping Club