Chrome DevTools Debugging

This skill explains how to use the Chrome DevTools MCP for browser debugging and UI verification.

Setup

Before using Chrome DevTools MCP, you must launch Chrome in headless mode with remote debugging enabled:

npm run chrome &

This runs Chrome with the required flags for Docker/containerized environments:

• --remote-debugging-port=9222 - Enables MCP connection
• --no-sandbox - Required for Docker
• --headless - Runs without display
• --disable-gpu - Avoids GPU issues in containers

Wait a few seconds for Chrome to start before using MCP tools.

Screenshots

IMPORTANT: Always save screenshots to the .screenshots directory using the filePath parameter:

mcp__chrome-devtools__take_screenshot with filePath: ".screenshots/descriptive-name.png"

Use descriptive filenames like:

• .screenshots/homepage-initial.png
• .screenshots/login-form-error.png
• .screenshots/after-click-submit.png

This allows debugging and visual comparison later.

Workflow Example

1. Start Chrome:

``bash npm run chrome & ``

Wait 2-3 seconds for startup.

1. Navigate to the dev server:

`` mcpchrome-devtoolsnavigate_page url="http://localhost:5173"; ``

1. Take a snapshot to see page structure:

`` mcpchrome-devtoolstake_snapshot ``

The snapshot shows element UIDs you can use for interactions.

1. Take a screenshot for visual verification:

`` mcpchrome-devtoolstake_screenshot filePath=".screenshots/page-state.png" ``

1. Interact with elements using UIDs from snapshot:

`` mcpchrome-devtoolsclick uid="14" mcpchrome-devtoolsfill uid="15" value="[email protected]" ``

Tips

• Prefer snapshots over screenshots - Snapshots are faster and provide element UIDs for interaction
• Always save screenshots locally - Use .screenshots/ directory with descriptive names
• Check if Chrome is running - If tools fail to connect, run npm run chrome & again
• Wait after navigation - Use wait_for to ensure page content loads before interacting