Claude Code can directly read and analyze image files as a multimodal LLM:
# Use the Read tool on PNG files - Claude sees the image directly
Read(file_path="/tmp/worldai-browser-evidence/screenshot.png")
# Claude will describe what it sees: user email, campaign list, UI state
python3 - <<'PY'from PIL import Image
import pytesseract
image_path = "/tmp/worldai-browser-evidence/screenshot.png"
img = Image.open(image_path)
text = pytesseract.image_to_string(img)
# Verify login successif"jleechan@gmail.com"in text or"jleechantest@gmail.com"in text:
print("SUCCESS: User is logged in")
if"My Campaigns"in text:
print("SUCCESS: Campaigns page visible")
print("\n=== Full OCR Output ===")
print(text)
PY
Example: Full Login Test
# 1. Navigate
mcp__chrome-superpower__use_browser(action="navigate", payload="https://mvp-site-app-s2-i6xf2p72ka-uc.a.run.app")
# 2. Click Sign In
mcp__chrome-superpower__use_browser(action="click", selector="//button[contains(text(), 'Sign in with Google')]")
# 3. Wait for OAuth popup (manual intervention if needed)# 4. Take screenshot
mcp__chrome-superpower__use_browser(action="screenshot", payload="login-test.png")
# 5. Extract and verify
mcp__chrome-superpower__use_browser(action="extract", payload="text", selector="body")
Playwright MCP Alternative
If Chrome Superpowers is unavailable, use Playwright MCP:
# Navigate
mcp__playwright-mcp__browser_navigate(url="https://mvp-site-app-s2-i6xf2p72ka-uc.a.run.app")
# Get snapshot
mcp__playwright-mcp__browser_snapshot()
# Click login button
mcp__playwright-mcp__browser_click(element="Sign in with Google button", ref="button-ref")
# Take screenshot
mcp__playwright-mcp__browser_take_screenshot()
Troubleshooting
Session Lost After Navigation
Firebase OAuth sessions may not persist across page navigations in MCP browsers.
Solution: Navigate to the target page directly after authentication completes, or use the same tab for all operations.
OAuth Popup Not Completing
The MCP browser cannot interact with popup windows.
Solution: Use a browser profile that's already logged into Google, or manually complete the OAuth flow when the popup appears.
"No token provided" Error
The API endpoint requires Firebase authentication.
Solution: Ensure you've completed the OAuth flow and the session is active. Check the header for your email address.
