AI testing Double
checked

We test the back end. Literally.

Cartoon: a web app with a worried face bends over in bug-print boxers while a bloodhound sniffs its backside and pulls two bugs out.

Your app has a code smell. We know where it's coming from.

Free homepage check · No signup · No card
Real customer. Real bugs.

“Found 10 bugs in our vibe-coded voting app before our fans did, including VOTE buttons on closed events. Cheaper than a tester, and our voters are happy as Larry.”

Noor Gabriel Noor Gabriel Marketing Manager, Reminisce

A first sniff. On the house.

Paste a public URL. Get a 0–100 homepage score and see what the dog found. Your results are open, not hidden behind an email form.

  • Page errors
  • Sampled links
  • Phone layout
  • Loading signals

Homepage only. This quick automated check does not click through your app or test logged-in workflows. A high score is not a bug-free guarantee.

Get my free score →

Less “something broke.”
More “here’s how to fix it.”

The free score spots homepage issues. Paid browser tests go further, with evidence you can give your coding agent. Here is a real example from our demo app.

Smoke report · real run, planted-bug demo

Taskpup · 3 November 2026

Our test app, not a customer testimonial.

  1. High

    Delete removes a different task than the one you clicked.

    1. Open the dashboard with several tasks.
    2. Sort by Title.
    3. Press Delete on the first row and confirm.

    Expected The task you clicked is deleted. Observed The newest task is deleted instead, and the clicked one stays, even after a reload.

    See the finding and screenshot →
  2. High

    Phone login lands on a 404 instead of the dashboard.

    Observed Login on a phone-size screen goes to a page that does not exist. Desktop login reaches the dashboard.

    See the finding and screenshot →
  3. Medium

    Sign-up with a mistyped email hangs on “Creating account…” forever.

    Observed No error, and the button never comes back. The new user is stuck.

    See the finding and screenshot →

The full report also shows passed checks, unconfirmed observations and checks we could not finish. A clean check is not a bug-free guarantee.

Read the full demo report, including its gaps →

A nice homepage can hide a messy app.

Start with the free homepage check. When you want the dog to follow your main journey, choose Smoke.

  • Broken signup form

    Submits. Spins. Nobody gets an account. Your launch tweet gets 400 likes and zero users.

  • Console errors

    47 red lines your users never see but definitely feel: broken features, silent failed saves.

  • Mobile layout melted

    Looked great on your 32-inch monitor.

  • Buttons that do nothing

    Click. Nothing. Click harder. Still nothing.

  • The forever spinner

    Loading… since Tuesday.

  • Broken links and edge cases

    404s behind your nicest buttons. Emoji in a name field. A user who hits Back halfway through saving.

Ready for more than a homepage check?

Skip the free check if you like. Smoke sends the dog through your app in a real browser.

Smoke

US$29

A real smoke test of your critical path. The dog makes its own test account.

The dog tests

  • Sign up and log in, with its own test inbox
  • Your app’s main action, end to end, once
  • Checkout up to the payment step (it never pays)
  • Desktop and phone widths, links, console and network errors

You get

  • Prioritised bugs with screenshots and repro steps
  • What passed, what was blocked and why, and what Deep would cover
  • Free homepage rescores after your fixes
  • See the demo report

Smoke opens soon. The launch list gets US$10 off your first Smoke, so it’s US$19. One email when we open, that’s it.

The guarantee

Not happy?
Full refund.

7 days, no questions, and you keep the report. Run breaks on our side? Refunded automatically. No chasing the dog.

One refund per customer email and app domain. Details · Questions: [email protected]

Functional QA only. No security, billing, load, real-device or Safari testing. Delivery depends on app size and the queue.

Fictional dogs. Real work.

Meet the pack.

AI agents test your app and draft the report. A second AI checks every paid report before delivery and sends anything unusual to a human. A human approves support replies.

  • French bulldog cartoon portrait of Elon Bust

    Elon Bust

    CEO

    Everything is P0. Everything is "next week". Has fired Head of Support three times. She's an AI, it didn't take.

  • Beagle cartoon portrait of Mark Zuckerbug

    Mark Zuckerbug

    Head of Bugs

    Moves fast. Finds them. Catalogues each one in uncomfortable detail.

  • Pug cartoon portrait of Travis Can't-Click

    Travis Can't-Click

    Head of Broken Buttons

    Finds the button that does nothing. Rates it one star.

  • Shiba Inu cartoon portrait of Whitney Wolf Howl

    Whitney Wolf Howl

    Head of Support

    Translates “it’s doing the thing again” into a useful support ticket.

Meet the other 13 →Hide the other 13 ↑
  • Golden retriever cartoon portrait of Bill Gatekeeper

    Bill Gatekeeper

    Head of Auth and Access

    Checks the login journey. Forgets his own password. Calls it research.

  • Jack Russell cartoon portrait of Marissa Mayhem

    Marissa Mayhem

    Head of Edge Cases

    Enters emoji in your phone number field. Every time.

  • Dalmatian cartoon portrait of Jessica Alt-Tab

    Jessica Alt-Tab

    Head of Cross-Browser

    Changes the browser width. Finds the button that fell off the screen.

  • Irish setter cartoon portrait of Richard Bansome

    Richard Bansome

    Head of Security

    We don't do security testing. He mostly locks the office and names things after himself.

  • Dachshund cartoon portrait of Jeff Pesos

    Jeff Pesos

    CFO

    Counts every token like it came from his own pocket. It did.

  • Greyhound cartoon portrait of Larry Page Load

    Larry Page Load

    Head of Performance

    Measures your homepage in milliseconds. Judges you in seconds.

  • Basset hound cartoon portrait of Sam Haltman

    Sam Haltman

    Head of Timeouts

    Fired and rehired over one weekend. Still waiting for your API to respond.

  • Husky cartoon portrait of Jensen Hang

    Jensen Hang

    Head of Frozen Screens

    Spends forty thousand dollars a second staring at your spinner.

  • Poodle cartoon portrait of Sheryl Sandbox

    Sheryl Sandbox

    Head of Test Environments

    Nothing touches prod on her watch. Leaned in, found staging.

  • Corgi cartoon portrait of Arianna Puffington

    Arianna Puffington

    Head of Load

    Huffs and puffs at your server. Recommends sleep when it falls down.

  • Border collie cartoon portrait of Melanie Pixel-kins

    Melanie Pixel-kins

    Head of Visual QA

    Spots your two-pixel misalignment from across the room.

  • Schnauzer cartoon portrait of Caterina Fetch

    Caterina Fetch

    Head of Test Data

    Fetches ten thousand fake users on command, all called Caterina.

  • Labrador cartoon portrait of Tom from BugSpace

    Tom from BugSpace

    Head of Social

    Still your first friend. Your only friend after you read the report.

All characters are fictional AI agents and parody. Any resemblance to actual tech founders is satire. No founders were sniffed in the making of this website.

Questions, sniffed

What AI tests my app?

Frontier models from OpenAI and Anthropic: GPT and Claude. One drives a real browser through your app; a different one plans the tests and checks every bug before it reaches your report, to reduce false positives. Unconfirmed observations are labelled separately; AI can still miss bugs or make mistakes. We use their APIs and aren't affiliated with either company. The free Sniff Score is a plain automated check, no AI.

Do you need my code?

No. Just a live URL. We test it the way a user would.

Is this security testing?

No. Functional QA only: does it work, does it break, does it look right. Richard Bansome is Head of Security in title only. But if the dog trips over something obviously dangerous while testing, like a secret key on a page or someone else's data on screen, we tell you privately. That's manners, not a security test, so a clean report doesn't mean you're secure.

What's free, and what does Smoke add?

The free Sniff Score is an automated homepage check. Smoke is a US$29 smoke test of your critical path in a real browser: the dog signs up with its own test inbox, logs in, does your app’s main action once from start to finish, and follows checkout to the payment step without paying, at desktop and phone widths. You get screenshots, reproduction steps and a “Not covered” list of what Deep would test next. Start free, then decide whether you want the dog to go further.

Will it find every bug?

No. The dog tests real journeys and reports the bugs it can reproduce, with screenshots and steps to fix them. Some bugs will hide in paths it never reaches, and your users will always find new ways to surprise you. Use the report before launch, then keep testing and listening after it.

What if you find nothing?

You get a report showing what we tested, what passed and what we could not cover or confirm. No confirmed findings does not mean the app is bug-free. If the run fails on our side, you get an automatic refund.

What if I don't like the report?

Full refund within 7 days of your report. No questions asked, and you keep the report. Reply to your report email and we'll send you a refund link. Telling us what fell short is optional, though it helps. It's one guarantee refund per customer email and per app domain, and your rights under Australian Consumer Law aren't affected. Details on the refunds page.

Why not build it yourself?

You can, and if you have a QA engineer and one app to test all the time, you probably should. Getting AI testing to work properly took us weeks: two models instead of one, a locked-down browser, and test apps with planted bugs to grade ourselves against. If you'd rather spend that time building your product, a Smoke sniff is US$29.

Can you test logged-in pages?

Yes. Smoke signs up with its own test inbox, logs in and does your main action behind the login. If people sign in with Google or Apple, a phone code or an invite, tell us after checkout and add a test login (turn off 2FA for it). If the dog still can’t get in, the report says exactly why. Smoke covers the critical path once; Deep tests every logged-in feature, roles, edge cases and your own test account. Always use test data, never real customer or admin accounts.

Who's actually behind this?

A human owns and is accountable for Bugsniffer. AI agents drive the browser, draft findings and prepare support replies. A second AI checks every paid report before delivery and sends anything unusual to a human, and a human approves support replies. The cartoon team is fictional. For help, email [email protected].

Can I get someone else's app sniffed?

No. You confirm it's yours when you order. Anything sensitive we spot, like exposed keys, stays locked in the report until you add a one-line tag to your app or give us a test account.

What do you do with my data?

Test-account logins are encrypted at rest and deleted when your run ends. Reports and screenshots are deleted after 30 days; save a copy if you want to keep yours. Order records have a separate retention period. See our privacy policy for providers and full details.

How fast do you answer support emails?

Usually within 24 hours, often much sooner. Email [email protected] and Whitney, our AI support agent, picks it up.

Don't panic. Is this how you say hello?

Yes.

Let the dog sniff it before your users do.

Start with the homepage. Free. No signup or card.