Video tutorial · The Metaverse Guy
Chrome DevTools MCP: Debug Web Apps Directly Inside Your AI Code Editor
Google’s new MCP server lets AI coding agents see live browser output, catch errors, and run performance audits—no more blind debugging.
- Published
- 5 October 2025
- Length
- 5 min
- Views
- 3,318 views
Google released the Chrome DevTools MCP server, which connects AI code editors like Cursor and Windsurf directly to a live Chrome browser. This removes the biggest limitation of AI coding agents: they couldn’t see what their generated code actually did. Now you can debug console and network errors, simulate user interactions, and run performance audits without leaving your editor.
The Blind Spot in AI Coding
Before this MCP, AI coding agents like those in Cursor or Windsurf had a major limitation: they couldn’t see what their generated code actually did in a browser. You’d get code, run it, find errors, and then have to manually feed console logs or screenshots back to the agent. It was a slow, blind debugging loop that broke the flow of vibe coding.
What Chrome DevTools MCP Brings
Google’s Chrome DevTools MCP server changes that. It connects your AI editor directly to a live Chrome browser, giving the agent real-time access to console errors, network requests, and page performance. You can now ask your agent to debug layout issues, simulate user clicks, or run a Lighthouse performance audit—all without leaving the editor.
Setting Up the MCP Server
I set it up in my editor by adding the MCP server configuration. The process is straightforward: you point the editor to the Chrome DevTools MCP server, and it automatically gains access to 26 tools for inspecting and interacting with the page. Once connected, I could see the browser tab being controlled by the agent, ready to respond to prompts.
Performance Testing with Prompts
To test it, I gave the agent a prompt to analyse the Largest Contentful Paint (LCP) of a page. The agent used the MCP to open the page, run a performance trace, and report back with specific timings and suggestions. I also compared the performance of web.dev with a sample landing page, and the agent quickly identified bottlenecks using the DevTools data.
Why It’s Better Than Playwright
Playwright is great for automated testing, but it operates outside the browser. Chrome DevTools MCP uses the actual DevTools protocol, so it can access performance audits, CSS layout details, and console errors in a way that feels native to the debugging workflow. For vibe coding, where you iterate quickly with an AI agent, this direct integration is far more effective.
Final Thoughts
This MCP removes the biggest friction in AI-assisted web development. Instead of context-switching between editor, browser, and DevTools, everything happens in one conversation. It’s a significant step forward, and I expect it to become a standard part of every web developer’s AI toolkit.
Tools used in this video
- Chrome DevTools MCP
- Cursor
- Windsurf
- KiloCode
- Augment Code
- Playwright
Questions
- What is Chrome DevTools MCP?
- It’s a server that connects AI code editors to Chrome DevTools, allowing AI agents to see live browser output, debug errors, and run performance audits directly from the editor.
- How do I set up Chrome DevTools MCP in Cursor?
- You add the MCP server configuration to your editor’s MCP settings, pointing to the Chrome DevTools MCP server. The video walks through the exact steps, and once connected, the editor can control a Chrome browser.
- What can Chrome DevTools MCP do that Playwright can’t?
- It uses the actual Chrome DevTools protocol, giving deeper access to performance audits, console errors, and layout debugging directly within the AI agent’s workflow, making it more seamless for vibe coding.
- Is Chrome DevTools MCP free?
- Yes, it is completely free to use.
Topics
Chapters
- 00:00 Google’s new AI update: Chrome DevTools MCP
- 00:23 Why AI agents had a big limitation before
- 01:00 How DevTools MCP solves the blind debugging problem
- 01:36 What this MCP can do (debug, simulate, performance audit)
- 02:15 Setting up the MCP server in your AI code editor
- 03:05 Exploring the 26 tools inside Chrome DevTools MCP
- 04:00 Testing with prompts (LCP & performance analysis)
- 05:20 Real example: Comparing web.dev with landing page
- 06:20 Why this is 10x better than Playwright
- 07:00 Final thoughts: Why this is a game-changer for vibe coding
Video description and links
This article was written from the video's description with the help of AI. Watch the video for the full walkthrough.



