**Short version:** Chrome DevTools MCP, from Google's Chrome DevTools team, lets a coding agent open a real Chrome and look at the same panels you would: performance, network and console. It is the fix for agents that change a page and then guess why it broke. It has about 52,000 GitHub stars, and version 1.9.0 shipped on 8 September 2026. Details below are from the project README.
Install it
Claude Code:
claude mcp add chrome-devtools -- npx -y chrome-devtools-mcp@latest --no-usage-statisticsOther clients: use command npx with args ["-y", "chrome-devtools-mcp@latest"]. You need a current Node.js LTS release and stable Chrome.
What it can do
- Performance: record a trace and pull out insights, so "the page feels slow" becomes a specific long task.
- Network: inspect requests and responses when an API call fails.
- Console: read errors with source-mapped stack traces.
- Automation: navigate, click and fill, with automatic waiting for results.
The flags worth knowing
- Privacy:
- --no-usage-statistics turns off Google's usage statistics, which are on by default.
- --no-performance-crux stops trace URLs going to the Chrome UX Report API.
- Environment:
- --headless for CI and remote machines.
- Context size:
- --slim exposes basic browser tasks only, which keeps tool definitions small.
Use a separate browser profile
The server exposes the browser's content to your AI client, which can inspect and modify any data in it. Never point it at a profile that is signed in to your bank, email or production admin tools. A throwaway profile or headless mode is the right default.
A workflow that shows its value
Ask the agent to fix a layout bug, then to open the page, read the console and confirm the fix before telling you it is done. That closes the loop that most agent-written frontend code leaves open. Pair it with the Playwright server from our best MCP servers list when you also need to reproduce multi-step flows.