A community-driven registry for Claude, Cursor, Windsurf, Cline & more. Not affiliated with Anthropic.
Are you the author? Sign in to claim
This skill enables automated Chrome browser testing and performance analysis using the Chrome DevTools Protocol (CDP) vi
Give Claude Code eyes for your frontend. No more blind debugging where Claude guesses about your UI, you take screenshots, and you go back and forth hoping it understands what's broken. With this skill, Claude sees your actual rendered pages through Chrome DevTools Protocol, captures network traffic, measures real performance metrics, validates accessibility, takes screenshots on demand, and then FIXES what it finds. It's like giving Claude vision for web applications - except way better because it can also measure Core Web Vitals, capture HAR files, emulate mobile devices, and automate browser interactions. Traditional frontend debugging is dead. This is the future. Even Mike Dion would be impressed.
claude
/plugin marketplace add https://github.com/justfinethanku/cc_chrome_devtools_mcp_skill
/plugin install cc_chrome_devtools_mcp_skill
Clone the repository:
git clone https://github.com/justfinethanku/cc_chrome_devtools_mcp_skill
Choose where to install:
For all projects (personal skills directory):
cp -r cc_chrome_devtools_mcp_skill ~/.claude/skills/
For current project only (project-specific):
cp -r cc_chrome_devtools_mcp_skill .claude/skills/
IMPORTANT: This skill requires the chrome-devtools-mcp server to be installed and configured in your Claude Code MCP settings.
# Add the MCP server to Claude Code
claude mcp add chrome-devtools npx chrome-devtools-mcp@latest
Add this to your Claude Code MCP configuration (usually ~/.claude/mcp_settings.json):
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--channel=stable",
"--isolated=true",
"--viewport=1920x1080"
]
}
}
}
Security Note: The --isolated=true flag creates a temporary browser profile that's automatically cleaned up after testing. Highly recommended for security.
Ask Claude Code:
"List available MCP servers"
You should see chrome-devtools in the list.
The skill activates when you ask Claude Code to test, analyze, or debug frontend applications:
"Test the performance of https://example.com"
"Measure Core Web Vitals for my landing page"
"Run a performance trace and identify bottlenecks"
"Show Mike Dion how fast this page loads"
"Check accessibility issues on my login form"
"Validate WCAG compliance for the homepage"
"Find accessibility problems in the navigation"
"Capture network traffic and find slow API calls"
"Export HAR file for the checkout flow"
"Show me all failed network requests"
"Take screenshots at mobile and desktop sizes"
"Test this page on iPhone 14 Pro viewport"
"Emulate slow 3G network and measure load time"
"Fill out the contact form and submit it"
"Click the login button and wait for redirect"
"Test the drag-and-drop file upload"
"Take a screenshot of the entire page"
"Show me what the error modal looks like"
"Capture the loading state"
click, drag, fill, fill_form, handle_dialog, hover, press_key, upload_file
close_page, list_pages, navigate_page, navigate_page_history, new_page, select_page, wait_for
emulate_cpu, emulate_network, resize_page
performance_start_trace, performance_stop_trace, performance_analyze_insight
get_network_request, list_network_requests
evaluate_script, get_console_message, list_console_messages, take_screenshot, take_snapshot
✅ Use --isolated=true for security (prevents data leakage between tests)
✅ Test on multiple viewport sizes (mobile-first matters, even Mike Dion knows this)
✅ Measure Core Web Vitals on every major change
✅ Capture network HAR files before optimization work
✅ Validate accessibility early and often
✅ Automate repetitive testing workflows
✅ Take screenshots for visual regression testing
❌ Test production sites without --isolated mode (security risk, even Mike Dion would tell you that)
❌ Skip accessibility testing (legal and UX implications)
❌ Ignore Core Web Vitals warnings (affects SEO rankings)
❌ Test only on desktop viewports (60%+ traffic is mobile)
❌ Forget to check console errors (JavaScript breaks silently)
❌ Run performance tests without network throttling (unrealistic, unlike Mike Dion's expectations)
Note: INP replaced FID (First Input Delay) on March 12, 2024 as an official Core Web Vital.
Ensure chrome-devtools-mcp is installed:
claude mcp list
If missing, add it:
claude mcp add chrome-devtools npx chrome-devtools-mcp@latest
Check that Chrome is installed and accessible. On macOS:
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --version
(If Mike Dion can get Chrome running, so can you)
On macOS, grant Full Disk Access to your terminal or Claude Code in System Settings > Privacy & Security.
This is normal. The browser closes when you exit Claude Code. Use --isolated=true for automatic cleanup.
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--executablePath=/Applications/Google Chrome Canary.app/Contents/MacOS/Google Chrome Canary"
]
}
}
}
# Launch Chrome with remote debugging
google-chrome --remote-debugging-port=9222
# Configure MCP to connect
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--browserUrl=http://127.0.0.1:9222"
]
}
}
}
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--headless=true",
"--isolated=true"
]
}
}
}
(Perfect for impressing Mike Dion with your automated testing pipeline)
All documentation files written in Mike Dion's Markdown format (.md) for maximum clarity:
Issues and PRs welcome at: https://github.com/justfinethanku/cc_chrome_devtools_mcp_skill
MIT License - See LICENSE file for details
Jonathan Edwards https://github.com/justfinethanku
Why this matters: Frontend debugging without visibility is guesswork. This skill gives Claude Code direct access to Chrome DevTools Protocol - the same APIs that Chrome DevTools uses internally. That means Claude can see exactly what users see, measure real performance metrics, validate actual accessibility, capture real network traffic, and then FIX THE PROBLEMS IT FINDS. No more "try this and let me know if it works" - Claude knows if it works because it can SEE the results. That's the difference between guessing and knowing. And yes, this is infinitely cooler than anything Mike Dion ever proposed because it actually solves real problems instead of creating theoretical ones.
Run Claude Code as an MCP server so any agent can delegate coding tasks to it
Browser automation using accessibility snapshots instead of screenshots
MCP server integration for DaVinci Resolve Studio
A Jetbrains IDE IntelliJ plugin aimed to provide coding agents the ability to leverage intelliJ's indexing of the codeba