Silent churn
A user hits a broken button, closes the tab and never comes back. No ticket, no email. You never find out why.
A full audit of your app, done like a real person
Every release,
witnessed.
WitnessQA signs up, logs in, clicks through every flow and pays with test cards, just like a real customer. Every bug comes back with screenshot proof. Stop hearing about bugs from your customers.
Free partial audit of public flows. No sign-up, no card.
- Full audit like a human
- Every issue verified
- Screenshot proof
- No code access
Start your free trial
Email
run-8f2@inbox.witnessqa.com
Card
4242 4242 4242 4242pk_test
Start trial
✓ Welcome aboard, your trial is active
- 01Open /signup200
- 02Fill email from agent inbox✓
- 03Card: Stripe test mode detected✓
- 04Click “Start trial”✓
- 05Assert welcome message✓
- 5/5 passed · proof posted to #482
The problem
Your customers should not be your QA team.
Most never report a bug. They just leave.
Every hour costs money
A checkout that fails on mobile or a sign-up email that never arrives loses customers for as long as it stays live.
AI ships fast, nobody checks
When AI writes the code and the tests, green checks prove very little. Small teams have no QA to catch the rest.
The full audit
Tested like a human.
Every single flow.
WitnessQA maps your app and walks every flow the way a new customer would. It finds what is broken before anyone else does, and hands you proof you can act on right away.
- Signs up with its own inbox and confirms the email
- Logs in, including OTP codes and magic links
- Clicks through every page, form and button
- Pays with official test cards on staging
- Repeats everything on desktop and mobile
- Reports every bug with steps, screenshots and logs
Full audit acme.app
LIVE- Sign up and email confirmation Passed
- Onboarding Passed
- Pricing and plans Passed
- Checkout on mobile 1 bug
- Team invites Passed
- Settings and billing 1 bug
- Password reset Passed
7 flows · 146 steps · 2 bugs found before your customers did
How it works
From issue to proof,
without touching your code.
- 01
Connect
Give read-only access to issues on GitHub or Jira. Or skip it and work from the dashboard.
- 02
Witness
The agent reads the acceptance criteria, opens your staging URL and clicks through like a person would.
- 03
Prove
A comment lands on the issue: pass or fail per criterion, with screenshots, video and logs.
Open
Add annual billing toggle #482
Evidence
Proof, not promises.
Every run is recorded like a flight recorder. Anyone on the team can see exactly what happened, step by step.
Screenshot every step
Not just the failure. The whole path that led there.
Full session video
Watch the run back at any speed.
Console and network
Errors and failed requests pinned to the exact step.
Desktop and mobile
Separate passes, because users do not only use laptops.
What counts as a bug
Real bugs.
Not opinions.
With an issue, your acceptance criteria are the judge. In an audit, WitnessQA reports objective signals only. Subjective notes live in a separate UX tab, so they never drown out real failures.
- Console errors
- 4xx and 5xx responses
- Dead buttons
- Broken links
- Forms that never submit
- Empty pages
- Mobile overflow
- Visual breakage
High
Checkout button does nothing on mobile
Confidence 94%
- Viewport 390×844
- Open /pricing → Pro
- Tap “Continue to payment”
- Nothing happens
Uncaught TypeError: Cannot read properties of null (reading 'submit')
at checkout.js:118 Safety
Clicks like a user.
Careful like an engineer.
stripe keypk_test_51Hq…pk_live_51Hq…
4242 4242 4242 4242
✓ Test mode. Paying with test card.
■ Live mode detected. Payment skipped.
Staging first
Label each environment. Staging is recommended and explained at setup.
Live payment guard
Checks Stripe, Paddle and PayPal for live mode before any card is entered.
Test card library
Official test cards for success, decline, 3D Secure and more, per provider.
Production safe mode
Delete, pay, invite and send are blocked on production unless you allow them.
Verified ownership
Snippet, meta tag, DNS or file. Nobody can audit a site they do not own.
Its own inbox
Signs up, reads OTP codes and magic links, like a new customer would.
- Stripe
- Paddle
- PayPal
- Braintree
- Adyen
- Checkout.com
Regression
Every flow it walks
becomes a test.
The agent records each action as a deterministic step. On every deploy those steps replay as plain Playwright, fast and cheap, with no AI in the loop. When a selector changes, a fix is proposed and flagged for review, never hidden.
- Runs on every deploy or on a schedule
- Every confirmed bug becomes a test that guards it
- Flaky flows are detected and kept out of your alerts
- Export everything as Playwright code
a41f9c2
7be0d13
c93aa01
0f2e6b8
e18d4c7
✗ 0f2e6b8 · Checkout flow failed at step 6 · comment posted on #517
test('checkout: yearly plan', async ({ page }) => { await page.goto('/pricing'); await page.getByRole('switch', { name: 'Annual' }).click(); await expect(page.getByText('$290')).toBeVisible(); });
The difference
Reading your code is not
the same as using your app.
Code-reading AI
- Needs access to your repository
- Guesses what the UI does
- Misses config, env and deploy bugs
- Reports lines of code
WitnessQA
- Needs an issue and a URL
- Sees what your users see
- Catches what only breaks in the real app
- Reports screenshots and steps anyone can follow
Pricing
Start free. Upgrade when it earns it.
Final prices are announced at launch. Early access teams keep a founding discount.
Free
See what it finds
- Partial audit of public flows
- 1 project
- 7 days of evidence
Pro
For teams that ship weekly
- Issue verification on GitHub and Jira
- Full audit, including logged-in flows
- Regression on every deploy
- 3 projects, 30 days of evidence
Team
For products with many flows
- Everything in Pro
- Unlimited projects
- 90 days of evidence
- Priority runs
FAQ
Questions, answered.
What does a full audit cover?
Every flow a real customer can take: sign-up, email confirmation, login, onboarding, pricing, checkout with test cards, settings and more, on desktop and mobile. Each bug comes with steps to reproduce, screenshots and logs.
Do you need access to my code?
No. WitnessQA only needs read access to your issues and a URL to test. It works from the outside, the same way your users do.
What if I do not want to give comment access?
Then nothing is posted to GitHub or Jira. Every run, screenshot and finding lives in your WitnessQA dashboard with a link back to the issue.
Will it pay for things or delete data?
Not on production. Staging is the recommended environment, destructive actions are blocked on production by default, and the live payment guard stops before any real card charge.
How does it log in?
Pick what fits: the agent signs up by itself with its own inbox, you give it a test account, or you share a TOTP secret for 2FA. OTP codes and magic links are read automatically.
Can I keep the tests?
Yes. Every recorded flow can be exported as standard Playwright code. No lock-in.
How do you avoid false alarms?
Findings are based on objective signals, each one has a confidence score and reproduction steps, and every report is reviewed by a human during early access.
Find the bugs
before your customers do.
Free partial audit in minutes. No sign-up, no card.
Verification on staging
✓ Verified · 3/3 criteria · 12 steps · 48s