---

name: webapp-testing description: Toolkit for testing and automating web applications. Prioritizes Claude Code Chrome integration for real browser testing. Falls back to Playwright scripts for CI/CD or headless scenarios. license: Complete terms in LICENSE.txt

Web Application Testing

Two approaches for testing web applications, with Chrome integration as the preferred method:

Decision Tree

CopyUser task → Is Claude Code running with --chrome flag?
    │
    ├─ Yes (Chrome available) → Use Chrome integration tools
    │       │
    │       ├─ Get tab context: tabs_context_mcp
    │       ├─ Create new tab: tabs_create_mcp
    │       ├─ Navigate: navigate tool
    │       ├─ Read page: read_page or find tools
    │       ├─ Interact: computer tool (click/type/screenshot)
    │       └─ Debug: read_console_messages, read_network_requests
    │
    └─ No (Chrome not available) → Fall back to Playwright
            │
            ├─ Static HTML? → Read file directly, write Playwright script
            └─ Dynamic app? → Use with_server.py + Playwright script

Chrome Integration (Preferred)

Prerequisites

• Google Chrome browser
• Claude in Chrome extension (v1.0.36+)
• Claude Code CLI (v2.0.73+)
• Start Claude Code with: claude --chrome

Core Pattern: Tab → Navigate → Read → Act

Copy1. Get tab context (required first step)
   → tabs_context_mcp

2. Create or select tab
   → tabs_create_mcp (for new tab)

3. Navigate to target
   → navigate tool with URL

4. Read page state
   → read_page (accessibility tree)
   → find (natural language element search)
   → get_page_text (raw text extraction)

5. Interact
   → computer tool: click, type, screenshot, scroll
   → form_input: fill form fields

6. Debug (if needed)
   → read_console_messages (filter with pattern)
   → read_network_requests (filter by URL pattern)

Example Workflows

Test Local Web App:

CopyNavigate to localhost:3000, try submitting the login form with
invalid data, check if error messages appear correctly.

Debug Console Errors:

CopyOpen the dashboard page and check the console for any errors
when the page loads. Filter for "Error" or "Warning".

Test Authenticated App:

CopyOpen my Google Sheet at docs.google.com/spreadsheets/d/abc123,
add a new row with today's date and "Test entry".

Record Demo GIF:

CopyRecord a GIF showing the checkout flow from cart to confirmation.

Chrome Tool Reference

Chrome Best Practices

1. Always get tab context first - Call tabs_context_mcp before other operations
2. Use fresh tabs - Create new tabs rather than reusing existing ones
3. Filter console output - Use pattern parameter to avoid verbose output
4. Handle blockers - Login pages, CAPTCHAs require manual intervention
5. Avoid modal dialogs - JavaScript alerts block browser events

Production Config Testing (CRITICAL)

Always test with production-like configuration, not local defaults.

Before testing, verify environment:

Copy# Check what config the app is using
grep -r "NEXT_PUBLIC_" .env*

# Start with production-like config
NEXT_PUBLIC_API_BASE="" NEXT_PUBLIC_WS_URL="wss://prod.example.com" npm run dev

During browser testing, verify in Network tab:

• ❌ No requests to localhost:8000 or localhost:3000 (indicates fallback)
• ❌ No unexpected 307 redirects (indicates trailing slash issues)
• ❌ No 401s on page load (indicates auth cascade)

Console checks:

Copy# Filter for fallback indicators
read_console_messages with pattern: "localhost|fallback|undefined"

Common failure patterns:

Chrome Limitations

• Requires visible browser window (no headless mode)
• Modal dialogs (alert/confirm/prompt) block further actions
• Not supported on Brave, Arc, or WSL

---

Playwright Fallback

Use when Chrome integration isn't available or for CI/CD pipelines.

Helper Scripts

• scripts/with_server.py - Manages server lifecycle

Always run with --help first to see usage. Treat as black-box scripts.

Server Lifecycle Pattern

Single server:

Copypython scripts/with_server.py --server "npm run dev" --port 5173 -- python test.py

Multiple servers:

Copypython scripts/with_server.py \
  --server "cd backend && python server.py" --port 3000 \
  --server "cd frontend && npm run dev" --port 5173 \
  -- python test.py

Playwright Script Template

Copyfrom playwright.sync_api import sync_playwright

with sync_playwright() as p:
    browser = p.chromium.launch(headless=True)
    page = browser.new_page()

    page.goto('http://localhost:5173')
    page.wait_for_load_state('networkidle')  # CRITICAL for JS apps

    # Reconnaissance
    page.screenshot(path='/tmp/inspect.png', full_page=True)
    buttons = page.locator('button').all()

    # Action
    page.click('text=Submit')

    browser.close()

Playwright Best Practices

• Always wait_for_load_state('networkidle') before inspecting dynamic apps
• Use sync_playwright() for synchronous scripts
• Always close browser when done
• Use descriptive selectors: text=, role=, CSS, or IDs

---

Comparison

Reference Files

• examples/chrome/ - Chrome integration patterns
• examples/playwright/ - Playwright script examples
• scripts/with_server.py - Server lifecycle helper