Vibe Test is an autonomous QA engineer that tests your running app — not your paperwork. It opens your app on its own, behaves like a real user, and comes back with evidence.
The problem
AI doubled how fast you ship. It didn't double your ability to test what you shipped — so bugs surface in production, tickets sit a week in “In Testing,” and a developer burns half a day reproducing a defect somebody else reported.
A bug ticket lands on your board.
2:00A bug ticket lands on your board.
It sits in the column. Everyone is asleep.
2:03Vibe Test opens your live app, reproduces it, captures the evidence.
Still sitting.
2:09Writes the fix, opens the PR, watches CI go green.
Still sitting.
2:14Re-tests, and moves the ticket to Done.
A developer finally starts reproducing it — half a day gone before the first line of the fix.
3 daysAlready shipped. Nobody woke up.
Nobody woke up. Nobody intervened.
You shipped at 2 AM. It was tested, fixed, merged and re-tested by 2:14.
An illustration of the Bugs Testing → Bugs Fixing path, against the wait it replaces. Real run times vary with your app.
Eleven services in two families. Pick one, press its Start and watch the run — then the thing it hands back.
Name what you're afraid of — “break checkout”, “find what's slow” — and Vibe Test hunts your live app for it until it gets there, or proves it can't.
New · Ask Vibe
Every service above has a launcher — and one shared door in front of all of them. Ask in plain words, point at anything in your project with @, and the chat comes back with a run to confirm. It proposes. You press.
Folded — and still answering
The chat at rest, on every tab. Close it mid-question and it keeps working in the bar — the step it is on, and a badge for what it noticed.
Open — proposing, then running
It doesn't only read and run. From the same conversation you can write— create a regression suite, a scenario or a journey, rewrite a story's acceptance criteria or a bug's write-up, tighten an AI target's contracts. Every one of those arrives as a card you approve, and every one can be undone.
Any of the eleven services ends the same way: not a red line in a log, but this.
A buyer with a valid card can complete every step up to payment, and then cannot pay.
The order is placed and a confirmation with an order number appears.
The spinner runs for 60s. No order is created; the network call to /api/payments never returns.
Severity, the service that found it, and the environment it happened in — so the first question in triage (“is this even real, and where?”) is already answered.
Written from what the run actually did, not inferred from a stack trace. A developer follows them instead of spending half a day rediscovering them.
The claim and the observation, side by side. This is the difference between “the test failed” and “the order never reached the payment provider”.
Captured when the finding is reported, not reconstructed later — the state of your app at the step that broke.
Before filing, it looks at your open bugs. A defect already on the board comes back linked to that ticket instead of opening a second one.
Jira, GitHub Issues, Flow OS — or the built-in board if you have none. The evidence goes with it.
Runs off your board too: Jira, GitHub Issues and Flow OS. No board at all? Vibe Board is built in.
The overview: Vibe Test opening a real app, testing it like a user, and reporting what it found with evidence.
Questions
Your app's URL. That is the setup — no SDK, no instrumentation, and no code access unless you want it. Repo access is optional, and only the services that write fixes actually require it.
Traditional automation runs scripts you wrote and breaks when your UI moves. Vibe Test opens your live app and behaves like a real user, and its regression locators heal themselves — your UI moves, your suite doesn't break. It also keeps a persistent quality memory per project, so the twentieth run is smarter than the first.
Yes — Jira, GitHub Issues and Flow OS. If you don't have a board at all, Vibe Test ships with Vibe Board, a built-in zero-config tracker.
Both. With Bugs Fixing enabled, a confirmed bug gets a written fix, an opened pull request and a watched CI run — then the ticket goes back for re-test automatically. The loop closes itself.
Every verdict ships with the evidence: screenshots, reproduction steps and a trail of what the run actually did. You don't have to take its word for it — you can audit it, or just ask it, right on the ticket.
It's autonomous, not unsupervised. Fix-cycle caps, self-echo guards, a pause switch and a hard rule against merging an empty diff are all built in, and your code and credentials stay encrypted.
Start your first sweep on your live app with one input: your URL.