Sybqa

Website QA for AI-built sites

Find what’s broken.
Before you share.

Built with Lovable, Bolt, v0 or Cursor? Test your site in a real browser. Get bugs with screenshots and steps you can reproduce.

Test my site for free Read a sample report

Free plan: 10 runs a month, no AI, no card. AI review from $19 a month.
Review the plan before it runs.

Recorded test fixture · signup

A bug you can reproduce.

Fail
Signup clears the name after an email error.

The form rejects an invalid email, then loses the name already entered.

Before · name entered
Captured signup fixture before submission: Name contains Example User.
After · name cleared
Captured signup fixture after submission: Name is empty and email validation says Enter a valid email address.
Steps to reproduce
  1. Open signupUse the purpose-built test fixture.
  2. Enter detailsAdd a name and an invalid email.
  3. Submit formThe email error appears; the name is cleared.
View sample report
Recorded evidence
Screenshots + browser replay
Fixture only
Purpose-built bug, not a customer result.

A clear path from question to evidence

From your link to a useful answer.

  1. Review the plan

    Share your link and the journeys that matter. Read the scope and limits before the browser starts.

  2. Watch the browser

    Follow real clicks, inputs and responses. See what happened at each step of the approved journey.

  3. Read the evidence

    Get screenshots and recorded steps where available. Anything blocked or not checked stays visible.

See the process

A closer look at a browser check.

Follow a scripted example, then inspect the measurements. The report separates confirmed failures from suggestions that need your review.

Example run on a made-up site

An animated example. Sybqa’s cursor opens a made-up plant shop called Fernly, clicks Login and lands on a 404 page, checks three AI leads, tries a bad email, then submits a valid one and waits five seconds with no response. The steps are listed as text below.

    Read this run as text
    1. 1 Open. Sybqa opens fernly.app, a made-up plant shop. It loads with a 200.
    2. 2 Login. It clicks Login in the menu and lands on a 404 page. Fail: “Login” goes to a 404.
    3. 3 Back. It goes back to the home page.
    4. 4 Look. The AI said the badge text is hard to read. The browser measured 7.1 : 1 contrast, so the lead is dropped. The AI said the password label looks cut off. The label reads “Pas” and there is nothing to measure, so it is left for you to review.
    5. 5 Email. It types “sam@”. The AI said the button text is hard to read. The browser measured 1.5 : 1 contrast. Fail. It presses Get Started and the form rejects the bad email. Pass.
    6. 6 Submit. It types sam@fernly.app and presses Get Started. Nothing happens for 5.0 seconds. Fail: signup never finishes.
    7. 7 Done. 6 results: 3 fail, 1 review, 1 pass, 1 dropped.
    Example run on Fernly, a made-up site. The steps are scripted for this page. Nothing here is a real customer or result. Run a free QA on my site →

    How to read the run

    1. The AI flags a possible problem.
    2. The browser measures it.
    3. You get a stamp.
    • PassMeasured and fine.
    • FailMeasured and broken.
    • ReviewNothing to measure, so you decide.
    • DroppedMeasured, and the AI was wrong.

    Measured facts. Review leads.

    AI can flag things that turn out to be fine. Sybqa checks measurable claims in the page and drops claims the measurements refute. Suggestions that can’t be measured stay marked for your review.

    Example AI findings and what the browser measured
    The AI saidThe browser measuredResult
    “The button text is hard to read.”1.5 : 1 contrastFail
    “The badge text is hard to read.”7.1 : 1 contrastDropped
    “The password label looks cut off.”Not measurableReview

    Why not just ask an AI?

    ChatGPT, Claude and the builders themselves can open your site now. Here’s what you get from each when you want to know whether it works.

    Sybqa compared with asking a chat AI or relying on your site builder’s own tests
    QuestionSybqaAsk ChatGPT or ClaudeYour builder’s own tests
    Who checks the workA separate testerThe same chat that suggests the fixesThe same agent that wrote the code
    SetupPaste a link and review the plan; test transactions need development keysTurn on a browser agent and write a prompt each timeBuilt in, and runs when the agent decides to
    What a finding meansBrowser evidence or a suggestion marked for review; refuted claims are droppedThe model’s opinionWhatever the agent chose to check
    Signups and paymentsTest accounts and payments with your development keys; cleanup attempted and leftovers reportedYou set them up and clean upDepends on the builder
    What you keepA report with screenshots, steps to reproduce, and JSON and JUnit filesA chat transcriptNotes in your build chat

    What you get

    Findings you can check

    Open the evidence available for each issue: screenshots, measurements or recorded steps. Model suggestions stay marked for review.

    A live view

    Follow recorded clicks, network requests and checks while the test runs.

    Signups and payments, in test mode

    Sybqa can sign up and pay with your test keys, then attempt refunds and account deletion. The report lists cleanup that needs your help.

    Your code and settings stay as they are

    Sybqa never needs your code. The free check submits nothing. Paid runs can use test accounts with your development key; cleanup attempts and leftovers appear in the report.

    Pricing, plainly

    Checks that don’t use AI are free. You only pay for the AI that looks for possible problems and writes up what it finds.

    Free

    Free

    $0

    • After you sign up: 10 runs a month.
    • Browser checks of links, forms, layout and contrast. No AI review.
    Create a free account

    AI review

    Sybqa Starter

    $19 / month

    • 15–29 full AI checks a month, depending on how big your site is.
    • AI suggests possible problems. Measurable claims are checked in the browser; other suggestions need your review.
    • Need more? Add 7–14 checks for $9, any time.
    • Cancel anytime in the billing portal.
    Start with Starter

    Full pricing details: what counts as a run, and free or paid

    Pricing questions

    How many checks do I get?

    Starter includes 500 AI credits a month. A full AI check on a typical site uses about 17 credits, and a large site up to 33, so that’s 15–29 checks. Checks that don’t use AI use no credits.

    What do I get without paying?

    after you sign up, 10 runs a month that don’t use AI. They check links, forms, layout and contrast.

    What happens when my credits run out?

    Buy a top-up in Billing: $9 adds 250 credits, about 7–14 checks. Top-up credits roll over while your plan is active. Or wait for your plan to renew. Runs that don’t use AI still work.

    Can I cancel?

    Yes. Cancel anytime in the billing portal. Your run history is kept.

    Catch broken flows before users do.

    Run a free QA on my site

    Common questions

    What is Sybqa?

    A QA tool for websites and web apps built with AI tools such as Lovable, Bolt, v0 and Cursor. It tests your site in a real browser and tells you what’s broken, with screenshots.

    Why not just ask ChatGPT, Claude or my builder to test it?

    Sybqa runs a separate browser check and records evidence. It measures claims about contrast, labels and target size; other model suggestions stay marked for your review. Start with a public link and review the plan. Test signups and payments need your development keys and test setup. See the comparison.

    Does Sybqa change anything on my site?

    Sybqa never edits your code or settings. The free check only reads your site and submits nothing. Paid runs can submit test forms or create throwaway accounts on test systems you configure. Payments require test-mode keys. Sybqa attempts cleanup afterwards and reports anything you need to remove manually.

    What’s in the report?

    Problems with the screenshots and recorded steps available for each check, a list of what couldn’t be tested, and JSON and JUnit files for your CI. Open the sample report.

    Do I need to connect my code?

    No. A URL is enough.

    Do I need a credit card to start?

    No. The free plan needs only an account, and you add a card only if you choose Starter. See what each plan includes.

    Can I run it on every deploy?

    Yes. Save a reviewed plan, create an API token, and add a GitHub Actions workflow. Each pull request or deploy gets a verdict and report link in the job summary. Read the CI guide.

    What is an AI credit?

    Credits measure the AI work in a check. A full AI check uses about 17 on a typical site and up to 33 on a large one. Checks that don’t use AI use none. See pricing.