Did the AI build
what you asked? TestRuna will test your app in a real browser to find out.
Paste its URL. TestRuna maps every feature, clicks through each one and shows you what works and what broke, with a screenshot for every step.
Free to start · Nothing to install · Your analytics stay clean
How to test your app with TestRuna
Paste your URL
Add the prompt you gave your AI builder if you have it. TestRuna opens the live app like a first-time visitor, follows its links and reads every page, form and button it can reach.
Get a test map
Every feature becomes a branch and every check a node. Your prompt turns into requirements, so you can see which of them the app really delivers and which it only claims.
Watch it light up
A real browser runs each test while you watch. Nodes turn green or red, and every step comes with a screenshot, the element it clicked and the reason it chose it.
What TestRuna checks for you
The questions a careful tester would ask, answered in minutes instead of an afternoon of clicking.
A test for every feature
Sign-up, search, cart, forms, settings: TestRuna designs tests from what it finds on up to 20 pages and from the requirements in your prompt. Add your own in plain English, such as “searching for MUG in capitals still finds the mug”.
A real browser, not a guess
Tests click, type and submit in Chromium, the way a person would, and you can watch every run live. Each step keeps a screenshot, so a pass or a fail is something you can see, not something you have to trust.
Evidence behind every result
A step only passes when the page shows proof, and our code checks that quote against the page. Failures get a second opinion from a different model, and when the evidence is unclear the test is marked for a human look instead of guessed.
Fix prompts for your AI builder
Every real failure becomes a precise request: where it happens, what happens now, what should happen and how to reproduce it. Paste it into Lovable, Bolt, v0 or Cursor, redeploy, and re-run only the tests that failed.
Recorded once, replayed in minutes
The first run records how each test gets through your app. Later runs replay those steps without waiting on AI, so checking the whole app after every change takes minutes, and when a page has changed, AI redoes just that step.
Safe on your live site
Tests only read by default: anything that signs up, posts or buys waits until you allow it, and nothing that deletes, pays or messages people ever runs on its own. Visits never reach your analytics, and test passwords are encrypted and never shown to the AI.
Why apps built with AI need testing
AI app builders are fast because they rewrite code with every prompt. That speed is also why they break things: a change to the checkout can quietly stop the cart from updating, and the builder will still tell you it is done. Nobody has time to click through every page after every prompt, so most of these regressions reach real users first.
TestRuna does that clicking for you. It keeps a map of what your app should do, runs it after each change and tells you exactly what stopped working, in words your AI builder understands. You spend your time fixing the two things that broke, not hunting for them.
Works where you build
In the workspace
Chat with TestRuna next to a live map of your app, ask why something failed and run the tests again with one click.
Inside your coding agent
Connect Claude Code, Codex, Cursor or Windsurf over MCP. Your agent deploys, asks TestRuna to test, reads the failures and fixes them.
In CI and GitHub Actions
Test your app after every deploy with the REST API, and fail the build when a test fails.
In your inbox
Send a report to you and your team after every run, or only when something breaks.
Start free, go Pro when you ship every week
100 free credits every month, enough to scan an app and run its tests a few times. Pro adds 2,500 credits a month from $39, and the same credits work in the workspace, over MCP and through the API.
See pricingQuestions
- Do I need to install anything?
- No. TestRuna runs in the cloud and can test your app through its public address, so there is nothing to add to your code. Paste the URL and the first results arrive in a few minutes.
- Which apps can it test?
- Any web app with a public URL, whether it was built with Lovable, Bolt, v0, Replit, Cursor or by hand, and hosted on Vercel, Netlify or anywhere else. Our browsers cannot reach localhost, so test a deployed or preview URL.
- Will it change my data or my analytics?
- Not unless you let it. Tests that create or edit data only run once you allow data changes, and we recommend doing that on a test or staging site. Analytics requests are blocked, so test visits never show up in your numbers.
- How accurate are the results?
- Every pass needs evidence from the page, checked by code, and every failure gets a second opinion from another model. When the page does not clearly show either, the test is marked for a human look. Screenshots of each step let you verify any result yourself.
- Can it test pages behind a login?
- Yes. Add a test account in the app's settings. Its password is encrypted, typed into your sign-in page by our browser and never shown to the AI. Signing in with Google or another provider, puzzles and two-factor codes still need a person.
- How is this different from writing Playwright tests?
- TestRuna runs on Playwright, but you never write a test or touch a selector. It designs the tests from your app and your prompt, judges every result from what the page shows and records the steps. Later runs replay those steps, and when your app changes, AI redoes only the step that no longer works.
- What does a failure report include?
- The step that failed, screenshots of the page, what it showed instead of the expected result, and whether a second model thinks it is a real bug or a problem with the test itself. Real bugs come with a fix prompt for your AI builder; problems like a missing test login are listed separately, so you never ask your builder to fix something that is not broken.
- What screen size do the tests use?
- Tests run in a desktop Chrome window of 1280 by 800 pixels, the layout most visitors on a laptop see. Phone layouts are not tested yet, so check those yourself or in your AI builder's preview.
- How much does it cost?
- Scanning an app costs 5 credits and each test run costs 1, while chat and fix prompts are free. The Free plan includes 100 credits a month, and Pro starts at $39 a month billed yearly. Tests that could not run because of a problem on our side are never charged.