| Exploration | Agents walk your live app, page by page | TestSprite calls your endpoints and records the real responses (Probe response and Raw call records on Test Coverage) |
| Plan grounding | Real screens and copy observed during exploration | Your API docs and instructions, plus the responses observed from your live API |
| Execution artifacts | Video recordings, step-by-step screenshots, console output | HTTP requests/responses, status codes, response bodies |
| Authentication | Sign-in to your app: test account credentials, OTP (one-time password) login, or SSO (single sign-on) such as Google | Credentials on the requests: a token, API key, or a refresh grant TestSprite renews before each run |
| Failure analysis | AI cause-and-fix with screenshot of the failing step | AI cause-and-fix with the offending request/response diff |
| Cleanup | Not applicable, no persistent resources created | Automatic post-run sweep of created records, files, and sessions |