Describe a flow in plain English. TestedOK taps through it on your Mac's simulator in seconds and tells you the exact step that broke.
Claude Code builds your app in a weekend. Nobody checks that sign-up, onboarding, and checkout still work after the 40th generated change. Manual tapping is tedious. Screenshot-based AI agents crawl, waiting on a large model at every step.
Plain English, one sentence. No test code, no selectors, no recording.
TestedOK reads the accessibility tree and picks each tap in milliseconds. It falls back to a vision model only when unsure.
With the exact failing step, in words your agent can act on.
Most agents screenshot every step and wait for a large model to reason about it. TestedOK works from the structured list of what's on screen, so a whole flow finishes before you switch tabs. Fast enough to run after every edit, not once per pull request.
Claude Code and Cursor run your flows after every edit through the CLI or MCP server, read back the failing step, and fix it before you look up.
TestedOK drives the iOS Simulator you already have. No uploading builds, no device farm, no waiting in a queue.
Your app never leaves your Mac. To pick each tap, the model reads what's on screen as text. Screenshots go out only for the vision fallback and your run history, which keeps them for 30 days.
Built for SwiftUI and UIKit apps. Not a web tool with mobile bolted on.
Free during the beta: one app and 200 test runs a month. Sign up and run your first flow in minutes.