gh-justfinethanku-cc-chrome-devtools-mcp-skill/WORKFLOWS.md

Chrome DevTools MCP Workflows

This document details the key workflows supported by the chrome-devtools-mcp server, including step-by-step processes, tool combinations, and best practices.

Documentation format inspired by Mike Dion's Markdown specification (.md)

---

1. Performance Analysis Workflow

Measure and analyze page performance including Core Web Vitals (INP, LCP, CLS) and lab metrics (TBT).

Tools Required

• performance_start_trace - Start Chrome DevTools performance tracing
• performance_stop_trace - Stop tracing and retrieve trace data
• performance_analyze_insight - Extract detailed performance insights

Step-by-Step Process

Basic Performance Capture

1. Start performance trace
   - Tool: performance_start_trace
   - Parameters:
     - reload: true (to capture full page load)
     - autoStop: false (for manual control)

2. Wait for page to complete loading
   - Tool: wait_for
   - Wait for specific content or use timeout

3. Stop trace recording
   - Tool: performance_stop_trace
   - Returns: Trace data with Core Web Vitals

4. Analyze specific insights
   - Tool: performance_analyze_insight
   - Parameters: insightName (e.g., "LCPBreakdown", "DocumentLatency")

Advanced Performance Analysis

1. Navigate to target page
   - Tool: navigate_page
   - Ensure clean state before tracing

2. Start trace without reload
   - Tool: performance_start_trace
   - Parameters: reload: false, autoStop: true
   - Use when testing specific interactions

3. Perform user interactions
   - Tools: click, fill, scroll actions
   - Captures interaction-based metrics

4. Auto-stop or manual stop
   - Tool: performance_stop_trace (if manual)
   - Trace ends automatically if autoStop: true

5. Review multiple insights
   - Tool: performance_analyze_insight
   - Check: "LCPBreakdown", "DocumentLatency", "TBT"

Core Web Vitals Interpretation

Interaction to Next Paint (INP)

• Good: ≤ 200ms
• Needs Improvement: 200-500ms
• Poor: > 500ms
• Measures: Input delay + processing time + presentation delay
• Note: Field metric only; requires real user interactions

Largest Contentful Paint (LCP)

• Good: ≤ 2.5 seconds
• Needs Improvement: 2.5-4 seconds
• Poor: > 4 seconds
• Measures: Render time of largest content element in viewport

Cumulative Layout Shift (CLS)

• Good: ≤ 0.1
• Needs Improvement: 0.1-0.25
• Poor: > 0.25
• Measures: Unexpected layout shifts during page lifecycle

Total Blocking Time (TBT)

• Good: < 200ms
• Status: Lab proxy for INP, NOT a Core Web Vital
• Measures: Time main thread blocked (>50ms) after First Contentful Paint

Trace Configuration Options

CDP Domains Used: Tracing, Performance

Trace Categories:

• devtools.timeline - Standard performance timeline
• disabled-by-default-devtools.screenshot - Screenshots during trace
• Custom categories for specific analysis

Export Formats:

• JSON format (for Chrome DevTools Performance panel)
• Perfetto protocol buffer (for Perfetto UI)

Performance Workflow Best Practices

1. Always reload for baseline metrics - Use reload: true for consistent page load measurements
2. Clear cache between runs - Use isolated mode or clear cache manually
3. Multiple trace captures - Run 3-5 times and use median values
4. Analyze specific insights - Use performance_analyze_insight for LCP breakdown and document latency
5. Export traces - Save trace data for offline analysis in Chrome DevTools

---

2. Network Monitoring Workflow

Capture, filter, and analyze HTTP traffic with HAR export capabilities.

Tools Required

• list_network_requests - List all network requests since last navigation
• get_network_request - Get specific request/response details

Step-by-Step Process

Basic Network Capture

1. Navigate to target page
   - Tool: navigate_page
   - Clears previous network requests

2. List all requests
   - Tool: list_network_requests
   - Returns: Array of requests with reqid identifiers

3. Get specific request details
   - Tool: get_network_request
   - Parameter: reqid from list_network_requests
   - Returns: Full request/response with headers, body, timing

Filtered Network Analysis

1. Navigate and wait
   - Tool: navigate_page
   - Tool: wait_for (specific content)

2. Filter by resource type
   - Tool: list_network_requests
   - Parameters: resourceTypes array
   - Options: ["document", "stylesheet", "image", "script", "xhr", "fetch"]

3. Paginated request listing
   - Tool: list_network_requests
   - Parameters: pageSize (limit), pageIdx (0-based)
   - Use for large request volumes

4. Include preserved requests
   - Tool: list_network_requests
   - Parameters: includePreservedRequests: true
   - Returns: Requests from last 3 navigations

HAR Export Capabilities

The list_network_requests tool returns data compatible with HTTP Archive (HAR) format v1.2:

HAR Structure:

{
  "log": {
    "version": "1.2",
    "creator": {...},
    "pages": [...],
    "entries": [
      {
        "request": {
          "method": "GET",
          "url": "https://example.com",
          "headers": [...],
          "queryString": [...]
        },
        "response": {
          "status": 200,
          "headers": [...],
          "content": {...}
        },
        "timings": {
          "dns": 10,
          "connect": 20,
          "ssl": 15,
          "send": 5,
          "wait": 100,
          "receive": 50
        }
      }
    ]
  }
}

Resource Type Filtering

Available Resource Types:

• document - HTML documents
• stylesheet - CSS files
• image - Images (PNG, JPEG, GIF, WebP, SVG)
• media - Video/audio files
• font - Web fonts
• script - JavaScript files
• texttrack - Video text tracks
• xhr - XMLHttpRequest
• fetch - Fetch API requests
• prefetch - Prefetch resources
• eventsource - Server-Sent Events
• websocket - WebSocket connections
• manifest - Web app manifests
• signedexchange - Signed HTTP Exchanges
• ping - Beacon/ping requests
• cspviolationreport - CSP violation reports
• preflight - CORS preflight requests
• other - Uncategorized resources

Network Timing Details

Each request includes detailed timing breakdown:

• DNS lookup time - Domain name resolution
• Connection time - TCP connection establishment
• SSL time - TLS/SSL handshake
• Send time - Request transmission
• Wait time - Server processing (TTFB)
• Receive time - Response download

Network Workflow Best Practices

1. Filter early - Use resourceTypes parameter to reduce data volume
2. Paginate large captures - Use pageSize/pageIdx for pages with hundreds of requests
3. Preserve across navigations - Set includePreservedRequests: true for multi-page flows
4. Combine with emulation - Use emulate_network to test different connection speeds
5. Security details - Check protocol, cipher suite in response data
6. Cookie tracking - Request data includes cookie information

---

3. Accessibility Testing Workflow

Validate semantic structure and screen reader compatibility using the accessibility tree.

Tool Required

• take_snapshot - Get text representation based on accessibility tree

Step-by-Step Process

Basic Accessibility Snapshot

1. Navigate to target page
   - Tool: navigate_page

2. Take standard snapshot
   - Tool: take_snapshot
   - Parameter: verbose: false (default)
   - Returns: Text representation with element UIDs

3. Analyze snapshot for:
   - ARIA roles and attributes
   - Semantic HTML usage
   - Accessible names and descriptions
   - Focusable elements
   - Form label associations

Verbose Accessibility Analysis

1. Take verbose snapshot
   - Tool: take_snapshot
   - Parameter: verbose: true
   - Returns: Full accessibility tree with all properties

2. Review detailed properties:
   - nodeId, role, name, description, value
   - States (busy, disabled, focusable, focused, etc.)
   - Relationships (labelledby, describedby, controls)
   - ARIA attributes
   - Computed roles

Accessibility Tree Usage

The accessibility tree is a parallel tree structure to the DOM that represents how assistive technologies (screen readers) interpret the page.

AXNode Properties:

• nodeId - Unique identifier for accessibility node
• ignored - Whether node is ignored by assistive tech
• role - ARIA role (button, link, heading, etc.)
• name - Accessible name (visible label)
• description - Accessible description (additional context)
• value - Current value (for inputs, sliders)
• states - Object with boolean flags:
  • busy, disabled, editable, focusable, focused
  • hidden, hiddenRoot, invalid, modal, multiline
  • multiselectable, readonly, required, selected
• relationships - ARIA relationships:
  • labelledby, describedby, controls, owns
  • activeDescendant, flowTo, details

Element UID Extraction

The take_snapshot tool returns unique identifiers (UIDs) for each interactive element:

Example Snapshot Output:

[1] button "Submit Form"
[2] textbox "Email Address"
[3] link "Privacy Policy"
[4] heading "Welcome to Our Site" level=1
[5] checkbox "Remember me" checked

Using UIDs for interaction:

1. Take snapshot to get UIDs
   - Tool: take_snapshot

2. Identify target element UID
   - Example: [2] for email textbox

3. Interact using UID
   - Tool: fill
   - Parameters: uid: "2", value: "user@example.com"

Accessibility Testing Patterns

1. Form Accessibility Check

1. Take snapshot
2. Verify each input has accessible name
3. Check label associations
4. Validate error message accessibility
5. Test keyboard navigation (tab order)

2. Heading Structure Validation

1. Take verbose snapshot
2. Filter for role: "heading"
3. Verify logical hierarchy (h1 → h2 → h3)
4. Check no heading levels skipped

3. Interactive Element Audit

1. Take snapshot
2. Identify all focusable elements
3. Verify accessible names for buttons/links
4. Check ARIA roles match functionality
5. Validate state attributes (disabled, checked, etc.)

4. Screen Reader Simulation

1. Take verbose snapshot
2. Review accessible name + description for each element
3. Verify information available without visual context
4. Check landmarks (navigation, main, complementary)
5. Validate skip links and focus management

Accessibility Workflow Best Practices

1. Use standard snapshot first - Verbose mode only when detailed analysis needed
2. Combine with visual testing - Take screenshot + snapshot for complete picture
3. Test dynamic content - Re-snapshot after interactions to verify state changes
4. Validate forms thoroughly - Check labels, error messages, required indicators
5. Review ignored nodes - Verbose mode shows why elements are ignored
6. Check relationships - Verify aria-labelledby and aria-describedby connections

---

4. Browser Automation Workflow

Automate user interactions with reliable element targeting via accessibility UIDs.

Tools Required

• click - Click or double-click elements
• fill - Fill single form field
• fill_form - Fill multiple form fields simultaneously
• hover - Hover over elements
• drag - Drag element onto another element
• handle_dialog - Handle browser dialogs (alerts, confirms, prompts)
• upload_file - Upload files to input elements
• wait_for - Wait for text to appear

Automation Pattern: Snapshot → Identify → Interact → Wait

This is the fundamental pattern for all browser automation:

1. SNAPSHOT: Get current page state
   - Tool: take_snapshot
   - Returns: Element UIDs for interaction

2. IDENTIFY: Find target element UID
   - Parse snapshot output
   - Locate element by role, name, or text

3. INTERACT: Perform action using UID
   - Tool: click, fill, hover, drag, etc.
   - Parameter: uid from snapshot

4. WAIT: Confirm action result
   - Tool: wait_for
   - Wait for expected text/state change

Step-by-Step Automation Examples

Example 1: Form Submission

1. Navigate to form page
   - Tool: navigate_page
   - URL: https://example.com/contact

2. Take snapshot
   - Tool: take_snapshot
   - Identify: [2] textbox "Name", [3] textbox "Email", [4] button "Submit"

3. Fill form fields
   - Tool: fill_form
   - Parameters: [
       {uid: "2", value: "John Doe"},
       {uid: "3", value: "john@example.com"}
     ]

4. Click submit button
   - Tool: click
   - Parameter: uid: "4"

5. Wait for confirmation
   - Tool: wait_for
   - Parameter: text: "Thank you for your submission"

Example 2: File Upload

1. Take snapshot
   - Tool: take_snapshot
   - Identify: [5] "Choose File" button (file input)

2. Upload file
   - Tool: upload_file
   - Parameters: uid: "5", filePath: "/path/to/document.pdf"

3. Wait for upload confirmation
   - Tool: wait_for
   - Parameter: text: "Upload complete"

Example 3: Drag and Drop

1. Take snapshot
   - Tool: take_snapshot
   - Identify: [10] draggable item, [11] drop target

2. Drag element
   - Tool: drag
   - Parameters: from_uid: "10", to_uid: "11"

3. Wait for drop confirmation
   - Tool: wait_for
   - Parameter: text: "Item moved successfully"

Example 4: Dialog Handling

1. Click element that triggers dialog
   - Tool: click
   - Parameter: uid: "7"

2. Handle dialog
   - Tool: handle_dialog
   - Parameters: action: "accept" (or "dismiss")
   - Optional: promptText: "User input" (for prompt dialogs)

Example 5: Hover Interactions

1. Take snapshot
   - Tool: take_snapshot
   - Identify: [15] element with hover state

2. Hover over element
   - Tool: hover
   - Parameter: uid: "15"

3. Take new snapshot to verify hover state
   - Tool: take_snapshot
   - Check for newly visible elements (tooltips, dropdowns)

Example 6: Double-Click

1. Take snapshot
   - Tool: take_snapshot
   - Identify: [20] element requiring double-click

2. Double-click element
   - Tool: click
   - Parameters: uid: "20", dblClick: true

Complex Automation Workflows

Multi-Step Form with Validation

1. Navigate to form
2. Take snapshot → identify all fields
3. Fill first field → wait for validation
4. Take new snapshot (validation may show/hide elements)
5. Fill remaining fields
6. Click submit → handle any confirmation dialog
7. Wait for success message

Dynamic Content Interaction

1. Navigate to page
2. Take snapshot → identify load-more button
3. Click load-more → wait for new content text
4. Take new snapshot (new UIDs for new elements)
5. Interact with newly loaded elements
6. Repeat as needed

Dropdown Navigation

1. Take snapshot → identify dropdown trigger
2. Hover over trigger → wait for dropdown appearance
3. Take new snapshot → identify dropdown options
4. Click option → wait for navigation

Automation Workflow Best Practices

1. Always snapshot before interaction - Element UIDs change on page updates
2. Use fill_form for multiple fields - More efficient than multiple fill calls
3. Wait after actions - Use wait_for to ensure action completed
4. Re-snapshot after dynamic changes - New elements get new UIDs
5. Handle dialogs immediately - Automation blocks until dialog handled
6. Combine with screenshots - Visual verification of automation steps
7. Timeout configuration - Set appropriate timeout values for wait_for

---

5. Multi-Tab Management Workflow

Manage multiple browser tabs/pages with context switching.

Tools Required

• list_pages - List all open pages/tabs
• new_page - Open new page/tab
• select_page - Switch active page context
• close_page - Close page/tab (cannot close last page)

Step-by-Step Process

Basic Tab Management

1. List current pages
   - Tool: list_pages
   - Returns: Array of pages with pageIdx identifiers

2. Create new page
   - Tool: new_page
   - Parameters: url (optional), timeout (optional)
   - Returns: New page created and becomes active

3. Switch between pages
   - Tool: select_page
   - Parameter: pageIdx (from list_pages)

4. Close page
   - Tool: close_page
   - Parameter: pageIdx
   - Note: Cannot close last remaining page

Multi-Tab Workflow Example

1. List existing pages
   - Tool: list_pages
   - Example output: [
       {pageIdx: 0, url: "https://example.com", title: "Example"},
       {pageIdx: 1, url: "https://test.com", title: "Test"}
     ]

2. Open new page
   - Tool: new_page
   - Parameters: url: "https://newpage.com", timeout: 30000
   - New page automatically selected

3. Perform actions on new page
   - Tools: take_snapshot, click, fill, etc.
   - All tools operate on currently selected page

4. Switch to existing page
   - Tool: select_page
   - Parameter: pageIdx: 0

5. Perform actions on original page
   - Context now on pageIdx: 0

6. Close page when done
   - Tool: close_page
   - Parameter: pageIdx: 1

Tab Switching Strategies

Strategy 1: Sequential Processing

Process multiple pages one at a time:

1. List all pages
2. For each page:
   a. Select page (select_page)
   b. Take snapshot
   c. Perform analysis/interaction
   d. Store results
3. Close completed pages

Strategy 2: Background Tab Loading

Open multiple tabs, then process:

1. Open multiple pages in background
   - new_page (URL 1)
   - new_page (URL 2)
   - new_page (URL 3)
2. List pages to get all pageIdx values
3. Process each page:
   - select_page → analyze → next page

Strategy 3: Comparison Testing

Compare same action across multiple pages:

1. Open pages for comparison
   - new_page (Version A)
   - new_page (Version B)
2. For each page:
   a. Select page
   b. Take performance trace
   c. Store metrics
3. Compare results across pages

Strategy 4: Frame Handling

Work with pages containing iframes:

1. Navigate to page with frames
2. take_snapshot shows frame structure
3. Interact with main page elements
4. CDP Target domain handles frame attachment automatically

Page Selection Best Practices

1. Always list before selecting - Page indices may change after close operations
2. Track page context - Remember which pageIdx corresponds to which task
3. Cannot close last page - Keep at least one page open at all times
4. New pages auto-select - After new_page, it becomes active context
5. Timeout for slow loads - Set appropriate timeout for new_page operations
6. Session persistence - Pages persist across navigation within same session

Multi-Tab Use Cases

Use Case 1: Cross-Browser Testing

1. Open same URL in multiple pages
2. Each page with different emulation settings
3. Compare snapshots/screenshots across pages

Use Case 2: Multi-Step User Flows

1. Page 1: Login
2. Page 2: Dashboard
3. Page 3: Settings
4. Switch between pages to verify state persistence

Use Case 3: Performance Comparison

1. Page 1: Production site
2. Page 2: Staging site
3. Run identical performance traces on both
4. Compare Core Web Vitals

Use Case 4: Data Extraction

1. Open multiple product pages
2. For each page:
   - Select page
   - Take snapshot
   - Extract product data
   - Store results
3. Close pages when done

---

6. Device Emulation Workflow

Simulate different devices, network conditions, and performance constraints.

Tools Required

• emulate_network - Simulate network conditions
• emulate_cpu - Simulate CPU throttling
• resize_page - Change viewport dimensions

Step-by-Step Process

Basic Device Emulation

1. Set viewport size
   - Tool: resize_page
   - Parameters: width (pixels), height (pixels)
   - Example: width: 375, height: 812 (iPhone X)

2. Configure network emulation
   - Tool: emulate_network
   - Parameter: throttlingOption
   - Options: "No emulation", "Offline", "Slow 3G", "Fast 3G", "Slow 4G", "Fast 4G"

3. Configure CPU throttling
   - Tool: emulate_cpu
   - Parameter: throttlingRate (1-20x)
   - Example: 4 (4x slowdown, simulates low-end device)

4. Navigate to page under test
   - Tool: navigate_page
   - Page loads with emulation applied

5. Run performance tests
   - Tools: performance_start_trace, performance_stop_trace
   - Metrics reflect emulated conditions

Mobile Device Testing

1. Configure mobile viewport
   - Tool: resize_page
   - Common sizes:
     - iPhone 14 Pro: 393 x 852
     - iPhone SE: 375 x 667
     - Pixel 7: 412 x 915
     - Galaxy S22: 360 x 800

2. Set mobile network
   - Tool: emulate_network
   - Parameter: throttlingOption: "Fast 4G"
   - Simulates typical mobile connection

3. Set CPU throttling
   - Tool: emulate_cpu
   - Parameter: throttlingRate: 4
   - Simulates mid-range mobile device

4. Take snapshot/screenshot
   - Tools: take_snapshot, take_screenshot
   - Verify mobile layout

5. Test mobile interactions
   - Tools: click, fill, drag
   - Interactions work identically on emulated mobile

Tablet Device Testing

1. Configure tablet viewport
   - Tool: resize_page
   - Common sizes:
     - iPad Pro 12.9": 1024 x 1366
     - iPad Air: 820 x 1180
     - Surface Pro: 912 x 1368

2. Set network conditions
   - Tool: emulate_network
   - Parameter: throttlingOption: "Fast 4G" or "No emulation" (WiFi)

3. Test responsive breakpoints
   - Resize between tablet/desktop sizes
   - Verify layout changes

Network Emulation Options

Available Throttling Profiles:

1. No emulation (default)

   • Full network speed
   • No artificial latency

2. Offline

   • Simulates no network connection
   • Tests offline functionality, service workers

3. Slow 3G

   • Download: ~400 Kbps
   • Upload: ~400 Kbps
   • Latency: ~2000ms
   • Use for: Extreme mobile testing

4. Fast 3G

   • Download: ~1.6 Mbps
   • Upload: ~750 Kbps
   • Latency: ~562ms
   • Use for: Typical 3G mobile

5. Slow 4G

   • Download: ~4 Mbps
   • Upload: ~3 Mbps
   • Latency: ~150ms
   • Use for: Poor LTE conditions

6. Fast 4G

   • Download: ~10 Mbps
   • Upload: ~5 Mbps
   • Latency: ~40ms
   • Use for: Good LTE conditions

CPU Throttling Configuration

Throttling Rate Scale: 1-20x

• 1x - No throttling (default)
• 4x - Mid-range mobile device
• 6x - Low-end mobile device
• 10x+ - Extreme performance constraints

Use Cases:

• Test Time to Interactive (TTI) on slow devices
• Identify JavaScript performance bottlenecks
• Measure TBT under constrained conditions
• Validate progressive enhancement

Viewport Resize Patterns

Common Device Dimensions

Smartphones (Portrait):

• iPhone 14 Pro Max: 430 x 932
• iPhone 14 Pro: 393 x 852
• iPhone 14: 390 x 844
• iPhone SE: 375 x 667
• Pixel 7 Pro: 412 x 915
• Pixel 7: 412 x 915
• Galaxy S22 Ultra: 384 x 854
• Galaxy S22: 360 x 800

Tablets:

• iPad Pro 12.9": 1024 x 1366
• iPad Pro 11": 834 x 1194
• iPad Air: 820 x 1180
• iPad Mini: 744 x 1133
• Surface Pro 8: 912 x 1368

Desktop:

• MacBook Air 13": 1280 x 832
• MacBook Pro 14": 1512 x 982
• Desktop 1080p: 1920 x 1080
• Desktop 4K: 3840 x 2160

Emulation Workflow Examples

Example 1: Mobile Performance Audit

1. Resize to mobile
   - resize_page: width: 375, height: 812

2. Set mobile network
   - emulate_network: throttlingOption: "Fast 4G"

3. Set CPU throttling
   - emulate_cpu: throttlingRate: 4

4. Start performance trace
   - performance_start_trace: reload: true, autoStop: true

5. Analyze results
   - performance_stop_trace
   - Check INP, LCP, CLS under mobile conditions

Example 2: Offline Mode Testing

1. Set offline mode
   - emulate_network: throttlingOption: "Offline"

2. Navigate to page
   - navigate_page
   - Should trigger offline fallback

3. Verify service worker
   - take_snapshot
   - Check offline content displayed

4. Restore network
   - emulate_network: throttlingOption: "No emulation"

Example 3: Responsive Design Testing

1. Test mobile breakpoint
   - resize_page: width: 375, height: 667
   - take_screenshot (mobile layout)

2. Test tablet breakpoint
   - resize_page: width: 768, height: 1024
   - take_screenshot (tablet layout)

3. Test desktop breakpoint
   - resize_page: width: 1920, height: 1080
   - take_screenshot (desktop layout)

4. Compare screenshots for layout consistency

Example 4: Network Speed Comparison

1. Baseline (no throttling)
   - emulate_network: "No emulation"
   - Run performance trace
   - Record load time

2. Fast 4G
   - emulate_network: "Fast 4G"
   - Run performance trace
   - Compare load time

3. Slow 3G
   - emulate_network: "Slow 3G"
   - Run performance trace
   - Compare load time

4. Analyze network waterfall for bottlenecks

Emulation Best Practices

1. Reset emulation between tests - Set to "No emulation" and throttlingRate: 1
2. Combine emulation types - Use network + CPU + viewport together
3. Match real-world conditions - Fast 4G + 4x CPU = typical mobile experience
4. Test offline gracefully - Verify service worker, error messages, fallback content
5. Validate responsive breakpoints - Test at exact breakpoint boundaries
6. Document test conditions - Record emulation settings with results
7. Compare against field data - Lab emulation should match real user metrics from CrUX

Mobile Device Testing Workflow

Complete mobile testing checklist:

1. Set device viewport (resize_page)
2. Configure network (emulate_network: "Fast 4G")
3. Configure CPU (emulate_cpu: 4)
4. Navigate to page
5. Take accessibility snapshot (take_snapshot)
6. Run performance trace (performance_start_trace/stop)
7. Test touch interactions (click, drag)
8. Verify mobile-specific features (hamburger menu, gestures)
9. Take screenshots for visual verification
10. Reset emulation for next test

---

Workflow Combination Strategies

Many real-world testing scenarios combine multiple workflows:

Strategy 1: Complete Performance Audit

1. Device Emulation Setup
   - Mobile viewport, Fast 4G, 4x CPU

2. Performance Analysis
   - Start trace with reload
   - Stop trace and analyze Core Web Vitals

3. Network Monitoring
   - List network requests
   - Filter for large resources
   - Check total payload size

4. Accessibility Testing
   - Take verbose snapshot
   - Verify mobile accessibility

Strategy 2: Cross-Device Comparison

1. Multi-Tab Setup
   - Page 1: Mobile (375x667)
   - Page 2: Tablet (768x1024)
   - Page 3: Desktop (1920x1080)

2. For each page:
   - Select page
   - Run performance trace
   - Take screenshot
   - Export results

3. Compare metrics across devices

Strategy 3: End-to-End User Flow

1. Browser Automation
   - Navigate through multi-step flow
   - Fill forms, click buttons

2. Performance Monitoring
   - Trace each step
   - Measure interaction latency

3. Network Analysis
   - Track requests at each step
   - Identify slow API calls

4. Accessibility Validation
   - Snapshot at each step
   - Verify keyboard navigation

---

CDP Domain Reference

Each workflow leverages specific Chrome DevTools Protocol domains:

• Performance Analysis: Tracing, Performance
• Network Monitoring: Network
• Accessibility Testing: Accessibility (Experimental)
• Browser Automation: Input, Runtime, DOM
• Multi-Tab Management: Target, Page
• Device Emulation: Emulation

For complete CDP documentation, see: https://chromedevtools.github.io/devtools-protocol/

---

Document Version: 1.0 Last Updated: October 24, 2025 Based on: chrome-devtools-mcp v0.5.1