| name | devtools |
| description | This skill helps launch and configure the Chrome DevTools MCP server, giving Claude visual access to a live browser for debugging and automation. Use when the user asks to set up browser debugging, launch Chrome with DevTools, configure chrome-devtools-mcp, see what my app looks like, take screenshots of my web application, check the browser console, debug console errors, inspect network requests, analyse API responses, measure Core Web Vitals or page performance, run a Lighthouse audit, test button clicks or form submissions, automate browser interactions, fill out forms programmatically, simulate user actions, emulate mobile devices or slow networks, capture DOM snapshots, execute JavaScript in the browser, or troubleshoot Chrome DevTools MCP connection issues. Supports Windows, Linux, and WSL2 environments. |
Chrome DevTools MCP Setup
Overview
This skill automates the setup and launch of Chrome with remote debugging for use with the chrome-devtools-mcp server. It handles environment detection, Chrome installation verification, MCP configuration, and browser launch across Windows, Linux, and WSL2.
GitHub Repository: https://github.com/ChromeDevTools/chrome-devtools-mcp
Why Give Claude Browser Access?
Without browser access, Claude is "coding blindfolded" - making changes without seeing the results. The Chrome DevTools MCP server provides 26 specialised tools that give Claude eyes into your running application:
| Category | Capabilities |
|---|---|
| Visual Inspection | Take screenshots, capture DOM snapshots, see rendered output |
| Console & Logging | Read console messages, catch JavaScript errors, debug issues |
| Network Analysis | Inspect API requests/responses, analyse headers, debug fetch calls |
| Performance | Record traces, measure Core Web Vitals (LCP, CLS, TBT), identify bottlenecks |
| User Simulation | Click elements, fill forms, drag-and-drop, handle dialogs |
| Device Emulation | Simulate mobile viewports, throttle CPU/network, test responsive design |
This enables Claude to:
- Verify changes visually instead of guessing if CSS/layout is correct
- Debug runtime errors by reading actual console output
- Test user flows by simulating clicks and form submissions
- Identify performance issues with real Core Web Vitals data
- Catch regressions by comparing screenshots before/after changes
Quick Start Workflow
Execute these steps in order:
Step 1: Detect Environment
Run the environment detection script to determine the platform:
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/detect_environment.sh
The script returns one of: windows, linux, or wsl2
Step 2: Verify Chrome Installation
Run the Chrome check script with the detected environment:
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/check_chrome.sh <environment>
The script outputs status:installed or status:not_installed. If Chrome is not installed, the script provides detailed installation instructions. See the Chrome Installation section below for manual installation options.
IMPORTANT: Do not proceed to later steps until Chrome is installed and verified.
Step 3: Check MCP Server Status
Verify if chrome-devtools-mcp is configured:
claude mcp list | grep -i chrome
If not installed, install with:
claude mcp add chrome-devtools -- npx chrome-devtools-mcp@latest --browserUrl http://127.0.0.1:9222
Step 4: Detect Running Dev Server
Check for running dev servers on common ports:
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/detect_dev_server.sh
This checks ports 5173, 5174, 5175, 3000, 3001, 8080, and 8000.
If no dev server is running and one is needed, offer to start it:
- For Vite projects:
npm run dev - For Next.js:
npm run devornpx next dev - For React CRA:
npm start
Step 5: Launch Chrome with Debugging
Launch Chrome with remote debugging enabled:
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/launch_chrome.sh <environment> <url> [headed]
Arguments:
<environment>:windows,linux, orwsl2<url>: Target URL (e.g.,http://localhost:5173)[headed]: Optional - passheadedfor visible browser, omit for headless (default)
Examples:
# Headless (default)
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/launch_chrome.sh wsl2 http://localhost:5173
# Headed (visible browser)
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/launch_chrome.sh wsl2 http://localhost:5173 headed
Platform-Specific Commands
WSL2 / Linux
# Headless
google-chrome --headless --remote-debugging-port=9222 --no-first-run --user-data-dir=/tmp/chrome-mcp http://localhost:5173 &
# Headed
google-chrome --remote-debugging-port=9222 --no-first-run --user-data-dir=/tmp/chrome-mcp http://localhost:5173 &
Windows (CMD/PowerShell)
REM Headless
start chrome.exe --headless --remote-debugging-port=9222 --no-first-run --user-data-dir=%TEMP%\chrome-mcp http://localhost:5173
REM Headed
start chrome.exe --remote-debugging-port=9222 --no-first-run --user-data-dir=%TEMP%\chrome-mcp http://localhost:5173
Chrome Installation
If Chrome is not detected, install it using one of these methods:
Linux / WSL2
Option 1: Direct download (recommended)
wget https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
sudo apt install -y ./google-chrome-stable_current_amd64.deb
Option 2: Add Google's repository
# Add signing key
wget -q -O - https://dl.google.com/linux/linux_signing_key.pub | sudo gpg --dearmor -o /usr/share/keyrings/google-chrome.gpg
# Add repository
echo 'deb [arch=amd64 signed-by=/usr/share/keyrings/google-chrome.gpg] http://dl.google.com/linux/chrome/deb/ stable main' | sudo tee /etc/apt/sources.list.d/google-chrome.list
# Install
sudo apt update
sudo apt install -y google-chrome-stable
Option 3: Chromium (open-source alternative)
sudo apt update
sudo apt install -y chromium-browser
Windows
Option 1: Download from Google Visit https://www.google.com/chrome/ and run the installer.
Option 2: Using winget
winget install Google.Chrome
Option 3: Using Chocolatey
choco install googlechrome
Option 4: PowerShell direct download
$installer = "$env:TEMP\chrome_installer.exe"
Invoke-WebRequest -Uri "https://dl.google.com/chrome/install/latest/chrome_installer.exe" -OutFile $installer
Start-Process -FilePath $installer -Args "/silent /install" -Wait
Remove-Item $installer
Verify Installation
After installation, verify with:
bash ${CLAUDE_PLUGIN_ROOT}/skills/devtools/scripts/check_chrome.sh <environment>
MCP Configuration
Quick Install
claude mcp add chrome-devtools -- npx chrome-devtools-mcp@latest --browserUrl http://127.0.0.1:9222
Configuration Reference
All flags can be passed via the args array in .mcp.json:
| Flag | Description | Default |
|---|---|---|
--browserUrl, -u |
Connect to running Chrome (e.g., http://127.0.0.1:9222) |
- |
--autoConnect |
Auto-connect to Chrome 145+ with remote debugging enabled | false |
--headless |
Run in headless (no UI) mode | false |
--isolated |
Use temporary user-data-dir, auto-cleaned on close | false |
--channel |
Chrome channel: stable, canary, beta, dev |
stable |
--viewport |
Initial viewport size (e.g., 1280x720, max 3840x2160 headless) |
- |
--executablePath, -e |
Path to custom Chrome executable | - |
--userDataDir |
Custom user data directory | ~/.cache/chrome-devtools-mcp/chrome-profile |
--wsEndpoint, -w |
WebSocket endpoint (alternative to --browserUrl) |
- |
--wsHeaders |
Custom WebSocket headers as JSON (use with --wsEndpoint) |
- |
--proxyServer |
Proxy server for Chrome | - |
--acceptInsecureCerts |
Ignore self-signed/expired certificate errors | false |
--chromeArg |
Additional Chrome launch arguments (array) | - |
--logFile |
Debug log file path (set DEBUG=* for verbose) |
- |
--categoryEmulation |
Include emulation tools | true |
--categoryPerformance |
Include performance tools | true |
--categoryNetwork |
Include network tools | true |
Basic Configuration
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--browserUrl",
"http://127.0.0.1:9222"
]
}
}
}
Headless with Isolated Profile
Best for CI/CD or automated testing - uses a temporary profile that's cleaned up automatically:
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--headless",
"--isolated"
]
}
}
}
Custom Viewport for Mobile Testing
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--browserUrl=http://127.0.0.1:9222",
"--viewport=390x844"
]
}
}
}
Using Chrome Canary/Beta
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--channel=canary",
"--headless",
"--isolated"
]
}
}
}
With Debug Logging
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--browserUrl=http://127.0.0.1:9222",
"--logFile=/tmp/chrome-devtools-mcp.log"
],
"env": {
"DEBUG": "*"
}
}
}
}
WebSocket Connection with Auth Headers
For connecting to remote Chrome instances with authentication:
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--wsEndpoint=ws://127.0.0.1:9222/devtools/browser/<id>",
"--wsHeaders={\"Authorization\":\"Bearer YOUR_TOKEN\"}"
]
}
}
}
To get the WebSocket endpoint, visit http://127.0.0.1:9222/json/version and look for webSocketDebuggerUrl.
Connection Methods
Method 1: Manual Connection (Recommended)
Start Chrome yourself with remote debugging, then connect via --browserUrl. This is the approach used in the Quick Start Workflow above.
When to use:
- Running Claude in a sandboxed environment
- Need full control over Chrome launch options
- Working with self-signed certificates
Method 2: Auto-Connect (Chrome 145+)
Let chrome-devtools-mcp automatically connect to a running Chrome instance.
Step 1: Enable remote debugging in Chrome:
- Navigate to
chrome://inspect/#remote-debugging - Follow the dialog to allow debugging connections
Step 2: Configure MCP with --autoConnect:
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": ["chrome-devtools-mcp@latest", "--autoConnect"]
}
}
}
When to use:
- Sharing state between manual testing and Claude-driven testing
- Avoiding WebDriver sign-in blocks (some sites block automated browsers)
- Want Chrome to prompt for permission before Claude connects
Method 3: Let MCP Launch Chrome
If you omit --browserUrl and --autoConnect, the MCP server will launch its own Chrome instance.
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": ["chrome-devtools-mcp@latest", "--headless", "--isolated"]
}
}
}
When to use:
- Fully automated workflows
- No need to maintain browser state
- CI/CD pipelines
User Data Directory
By default, chrome-devtools-mcp uses a persistent profile at:
- Linux/macOS:
$HOME/.cache/chrome-devtools-mcp/chrome-profile-$CHANNEL - Windows:
%HOMEPATH%/.cache/chrome-devtools-mcp/chrome-profile-$CHANNEL
This profile is shared across all MCP sessions, preserving cookies, local storage, and login state.
Use --isolated for a fresh, temporary profile that's automatically cleaned up when the browser closes.
Troubleshooting
For detailed troubleshooting steps, read references/troubleshooting.md.
Quick Checks
Test MCP server runs:
npx chrome-devtools-mcp@latest --helpVerify Chrome is listening:
curl -s http://127.0.0.1:9222/json/versionCheck for existing Chrome processes:
# Linux/WSL2 pgrep -a chrome # Windows tasklist | findstr chrome
Common Issues
| Issue | Solution |
|---|---|
| "Target closed" error | Close all Chrome instances, restart with debugging |
| Module not found | Clear npm cache: rm -rf ~/.npm/_npx && npm cache clean --force |
| Connection refused | Ensure Chrome launched with --remote-debugging-port=9222 |
| Port already in use | Kill existing Chrome or use different port |
| Chrome won't start in sandbox | Use --browserUrl to connect to manually-started Chrome |
| WebDriver sign-in blocked | Use --autoConnect to connect to your normal browser session |
| VM-to-host connection fails | See references/troubleshooting.md for port forwarding guidance |
Known Limitations
Operating System Sandboxes
Some MCP clients sandbox the server using macOS Seatbelt or Linux containers. In sandboxed environments, chrome-devtools-mcp cannot start Chrome (which requires its own sandbox permissions).
Workarounds:
- Disable sandboxing for chrome-devtools-mcp in your MCP client
- Use
--browserUrlto connect to a Chrome instance started outside the sandbox
Security Considerations
The remote debugging port exposes your browser to any application on your machine. When debugging is enabled:
- Avoid browsing sensitive sites (banking, email with sensitive data)
- Use
--isolatedfor a separate profile - Close Chrome when done debugging
Verification
After setup, verify the connection works:
- Chrome should be running with remote debugging
- MCP server should connect to
http://127.0.0.1:9222 - Test with
mcp__chrome-devtools__list_pagestool