Skip to main content
Every time you run a test (manually or scheduled), Propolis creates a test run record showing exactly what happened during execution.

Understanding Test Runs

The Test Runs page shows a table of all executions: Name - Which test ran Status - Passed, Failed, Running, Queued, Stopped, or Agent Error Suite - The test suite it belongs to Batch - If part of a scheduled run Time - When it executed Click any row to open the detail panel.

What’s in a Test Run

When you open a test run, you see: Video - Watch exactly what happened. See every page visited, what the agent clicked, and what the browser displayed. Timeline - Every action the agent took with screenshots, reasoning, and timing. Console & Network - Browser logs and HTTP requests for debugging (see below). Actions - View Definition, Delete, or Flag Error.

When a Test Fails

Test Caught a Bug

Video shows your app broken? The test is working correctly.Next steps:
  • Share the test run with your team
  • Fix the bug in your app
  • Re-run to verify

Test Needs Fixing

App working but test failed? The test needs updating.Next steps:
  • Use debugging tips below
  • Update test steps
  • Re-run to verify

Debugging Broken Tests

Check:
  • Video - Was the button/link visible on screen? Was page fully loaded?
  • Timeline - What was the agent looking for?
  • Console - Any JavaScript errors preventing the page from loading?
Fix:
  • Make AI step goal more specific (include color, position, text on the button)
  • Add wait step before the interaction
  • Check if content is in a modal or section that needs to be opened first
Check:
  • Video - Did page redirect to login?
  • Network - Look for 401/403 response codes in API calls
  • Console - Any auth errors logged?
Fix:
  • Update browser profile with fresh login
  • Verify profile is selected in test configuration
  • Check if session expired during execution
Check:
  • Video - What did the page actually show?
  • Timeline - What was being verified?
  • Console - Any errors affecting page content?
Fix:
  • Make check prompt less strict
  • Ensure previous steps completed successfully
  • Add wait step if content loads dynamically
Check:
  • Network - Look for slow API calls (check request duration)
  • Video - What was loading when it timed out?
  • Timeline - Which step took too long?
Fix:
  • Add wait steps for slow-loading content
  • Check for network issues in test environment
  • Verify API endpoints are responding quickly