Google AI introduces the public preview of Chrome Devtools MCP: Make your coding agent control and check the real-time Chrome browser

Google released Public preview of “Chrome Devtools MCP“,” Model Context Protocol (MCP) server, which enables AI encoding agents to control and check real Chrome instances – track performance, check DOM and CSS, execute JavaScript, read console output, and automate user flow. Start limits on code that directly targets code: They usually cannot be inspired by gradually building frequent people or step by step. With MCP, Google turns the static suggestion engine into a loop-enclosed debugger, running measured values ​​in the browser before filing a fix.

What is Chrome DevTools MCP?

MCP is an open protocol for connecting LLM to tools and data. Google’s DevTools MCP acts as a professional server that exposes Chrome’s debug surface to MCP-compatible clients. Google’s developer blog positioned it as “bringing the power of Chrome DevTools to the AI ​​coding assistant” and in specific workflows such as startup performance tracking (e.g. performance_start_trace) Target URLs, then let the agent analyze the result traces to propose optimizations (e.g., diagnose the largest content paint).

Functional and tool surfaces

The official GitHub repository records an extensive tool set. Beyond performance tracking (performance_start_trace,,,,, performance_stop_trace,,,,, performance_analyze_insight), the agent can run navigation primitives (navigate_page,,,,, new_page,,,,, wait_for), simulate user input (click,,,,, fill,,,,, drag,,,,, hover) and ask for runtime status (list_console_messages,,,,, evaluate_script,,,,, list_network_requests,,,,, get_network_request). Screenshot and Snapshot utility provides visual and DOM-State capture to support differences and regression. The server uses Puppeteer under the hood for reliable automation and wait semantics and speaks to Chrome via the Chrome DevTools protocol (CDP).

Install

For MCP customers, the settings are intentionally minimal. Google recommends adding a single configuration section that will npxalways track the latest server builds:

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": ("chrome-devtools-mcp@latest")
    }
  }
}

The server integrates with multiple proxy front-ends: MCP support for Gemini CLI, Claude Code, Cursor and Github Copilot. For VS code/adverb, repo documentation code --add-mcp Single line for Claude code, claude mcp add Same command npx Target. Package targets Node.js≥22 and current chromium.

Sample Agent Workflow

Google’s announcement highlights pragmatic tips that demonstrate end-to-end loops: validating proposed fixes in a live browser; analyzing network failures (e.g., CORS or blocking image requests); simulating user behavior, such as submission form to copy errors; checking for layout issues by reading DOM/CSS in context; and running automatic performance audits to reduce LCP and other core network vitality. These are all operational agents that can now be verified by actual measurements rather than heuristics.

Summary

The public preview of Chrome DevTools MCP is a practical inflection point for proxy front-end tools: it is based on AI assistants in real browser telemetry (performance trajectory, DOM/CSS state, network, network and console data), so the recommendation is driven by measurements rather than guesswork. First-party servers shipped by Chrome DevTools team are available through npx And target customers with MCP and use Chrome/CDP under the hood. Regression and sheet UI flow are expected, as well as more stringent validation of performance work.

Check Technical details and Github page. Check out ours anytime Tutorials, codes and notebooks for github pages. Also, please stay tuned for us twitter And don’t forget to join us 100K+ ml reddit And subscribe Our newsletter.

Talk to us about content partnerships/promotions on Marktechpost.com

Asif Razzaq is CEO of Marktechpost Media Inc. As a visionary entrepreneur and engineer, ASIF is committed to harnessing the potential of artificial intelligence to achieve social benefits. His recent effort is to launch Marktechpost, an artificial intelligence media platform that has an in-depth coverage of machine learning and deep learning news that can sound both technically, both through technical voices and be understood by a wide audience. The platform has over 2 million views per month, demonstrating its popularity among its audience.