BLOG
BlogTopic

Testing apps built with AI

For apps written by Claude Code, Cursor, Lovable, Bolt or Replit: checking that what the agent called done works for a user.

GuidesWebTesting an app you built with AI: a guide for foundersYou built it with Claude Code, Cursor, Lovable or Bolt and it looks done. How to check it works for a stranger and keeps working after every change.2 min read GuidesWebYour coding agent said “done”. Here is what it did not checkAgents mark features complete after the code compiles and the unit tests pass. Why that misses broken screens, and how to check the feature as a user would.3 min read GuidesWebSign-up email links to localhost? Catch it before your users doA confirmation or reset email that points at localhost or an old preview is a common bug in AI-built apps. A test that opens the real email finds it on the first run.3 min read GuidesWebIt works in preview and breaks in production. Test the live addressAI app builders blur preview and production, so the version you clicked through is not the one users get. Test the deployed address after every deploy instead.2 min read GuidesWebFix one bug without breaking another: one bug, one testAI-built apps fall into a loop where each fix breaks something that worked. Keep a saved test for every bug you fix and run them all after each change.2 min read GuidesWebTurn a failed test into a fix prompt for your coding agentDescribing a bug to an AI agent is where fixes go wrong. Copy the fix prompt from a failed run instead: the check, what was seen, the steps and the proof.2 min read GuidesWebLet Cursor check its own work in a real browserConnect Testsy to Cursor over MCP so the agent can save a plain-language test, run it against your deployed app or a phone build, and read the verdict.3 min read GuidesWebLet Codex check its own work in a real browserAdd Testsy to Codex with one command. The agent drafts a plain-language test, runs it against your app, reads every check and reruns after its fix.2 min read GuidesWebTest from Gemini CLI, the Claude desktop app and claude.aiInstall Testsy as a Gemini CLI extension or a Claude connector, then ask in plain words for a browser or phone test and get a verdict with evidence.2 min read GuidesWebTest a Lovable, Bolt or Replit app by its address, no code neededBuilt your app in Lovable, Bolt, Replit or v0? Paste the published address, describe a user's journey in plain words, and get a verdict with a replay.2 min read GuidesWebSaved, then gone after a refresh: test that your app really savesA screen can show data that was never stored. One plain-language test creates a record, reloads the page and checks it is still there, the bug AI-built apps ship most.2 min read GuidesWebTest that one user cannot see another user's dataApps built with AI often show one account's records to another. A two-person test signs in as each, in separate browsers, and checks what each one can see.2 min read GuidesWebTest password reset from the email to the new passwordA reset email that links to the wrong host, or a link that fails in a new browser, locks users out. One test requests the reset, follows the link and signs in again.2 min read GuidesWebCatch the blank white screen after a deployThe most reported failure of AI-built apps is a page that renders nothing in production. A test that reads one value and records the console catches it on the first run.2 min read GuidesWebTen tests to run before you share your AI-built appA pre-launch checklist for apps built with Claude Code, Cursor, Lovable or Bolt, written as ten plain-language tests you can run on the live address today.2 min read GuidesWebWrite a test without writing code: steps, checks and valuesA plain-language test is only as good as its checks. How to write steps a tester can follow and checks it can judge, with examples that pass review first time.2 min read GuidesWebWhat an AI tester cannot check for youA browser test proves what a user can see and do. It does not audit security, load or code. An honest list of what Testsy covers, and what still needs a person.2 min read GuidesWebCheck AI-written code before your customers doTeams ship more code than anyone reviews line by line. Checking each deploy from the outside, as a customer would, catches what review and unit tests miss.2 min read GuidesiPhone and iPadAndroidLet your coding agent test the mobile app it just builtClaude Code, Cursor or Codex builds the app, uploads the build over MCP without an API key in the shell, runs your saved app tests and reads the verdict.2 min read GuidesWebSign-up emails that stop arriving after launchSupabase's built-in mailer sends 2 emails an hour, only to your own team. Sign-up works for you and fails for everyone else. A test with its own mailbox catches it.2 min read GuidesWebPaid, but still on the free plan: test the Stripe webhookStripe takes the money, the app never hears about it. A test that pays in test mode and checks the plan changed catches it.2 min read GuidesWebUploads that work for you and fail for your usersA photo upload works when you try it and fails for a normal user with a 403 from storage rules. Test it as a normal account, with a real file, and check after a reload.2 min read GuidesWebTest your app as a brand-new user, not as yourselfYour account has data, a team and every flag on. A new user's has nothing. Sign up fresh on every run and check the first thing a new user must be able to create.2 min read GuidesWebTest what users see when a request failsApps built with AI handle the happy path and show a blank page or a spinner forever when an API fails. Test the error message, the Retry button and the slow case.3 min read GuidesWebSign in with Google works locally and fails in production“Error 400: redirect_uri_mismatch” is what new users see when the Google sign-in was only set up for localhost. A test presses the button and checks where it lands.2 min read