TesterArmy, a Y Combinator startup, has launched a hosted service that runs an app's critical user journeys from instructions written in plain English. You describe the flow, its AI agent navigates the pages, fills the forms, and reports what broke, with no SDK, no test scripts and nothing added to your repo.
The detail that makes it more than a demo is the login handling. The agent signs in with stored credentials, completes OAuth, and receives one-time passwords through dedicated per-agent inboxes, with credentials held encrypted at rest using AES-256-GCM. That is the part most browser-automation tools quietly skip, because OTP and OAuth are where scripted tests usually die.
It runs continuously and alerts the team when a key flow fails, slotting into Slack, GitHub and CI/CD. The open question is whether teams will hand a hosted agent standing access to production logins and the OTPs that protect them, which is exactly the trust a service like this needs to earn.