For online shops, websites, web apps, MCP and APIs

Find the issues
that cost you sales.

Your site, shop, web app, MCP and API, tracked daily. A signal the moment something breaks.

Customers came to buy, not to report issues. Know before they do and fix it fast, so nobody leaves over something that does not work.

Free, no sign-up, about 5 minutes. The agent walks your public flows on a desktop and a phone, and you watch each flow as it goes. After you sign up it also signs in and tests every flow, including the pages behind the login.

Want the full thing now? Start the 7-day trial for $1. Why $1? We use it to verify that you are a real user. Cancel within 7 days and you pay nothing more.

Pages 0.1 s faster lifted retail conversions 8.4% (Deloitte & Google). 1 in 6 abandoned checkouts is a site error (Baymard).

  • Shopify, WooCommerce and any site
  • Tracked daily, and after every deploy
  • Know before your customers do
  • Every path to the order
  • Checkout with test cards
  • Behind the login too
  • Phone and desktop
  • Speed measured on a phone
  • SEO as Google and AI read it
  • Found by ChatGPT, Claude and Perplexity
  • HTTPS and security policy checked
  • Your API and MCP endpoints checked
  • Track every issue on one board
  • Full history of every issue and audit
  • API and MCP for your own agents
  • We suggest a fix for you
  • Cancel anytime

A week at a made-up shop. In your report, the pages, the issues and the fixes are yours.

What you get

One place.
Eight ways to put it to work.

Start with a free audit, or let the agent track your site, shop, web app, MCP and API daily. Everything it finds comes with a screenshot and a suggested fix.

Free audit

Try it on your public pages

3 to 5 public flows on a phone and a desktop, with a screenshot of every step. No sign-up, no card.

Run a free audit

Full audit

Every path to the order

Product, search, cart, discount codes, checkout with test cards, sign-up, login and account, on a phone and a desktop.

Start the 7-day trial for $1

Track daily

Tracked daily, and after every change

Something breaks on Tuesday, you know on Tuesday, before your customers do. The next run confirms your fix.

See a week

AI visibility

Found by Google, ChatGPT and Claude

Which AI crawlers get in, which pages they can read, and whether you have an llms.txt.

See the check

Beyond the pages

Speed, HTTPS and your endpoints

Speed on a phone, HTTPS and security policy, dead links, and your API and MCP endpoints.

What the audit checks

Issue board

What got fixed, what is open

Every issue is a card with its proof: the screenshot, the steps and a suggested fix.

What counts as an issue

Your own agents

API and MCP

Add WitnessQA to Claude Code, Cursor or any MCP client, or call the API. Your agent gets every problem back with a suggested fix.

Read the docs

We fix it for you

No developer at hand?

We can fix it for you: you pay per fix, it is fixed once and stays yours for good, and if it breaks again we fix it again free. No subscription.

Book a 30-minute call

The problem

Customers do not report problems.
They just leave.

Broken

A dead size picker, a failing discount code, a checkout that breaks on phones. No ticket, no email. The order is gone.

Slow

Every extra second on a phone costs orders. Speed tools list warnings, not what to change.

Invisible

A page left on noindex, a sitemap that 404s. Google and AI assistants skip you, and nobody notices.

The report

One link. Three answers.
What to fix first.

One shareable report: how well the site works, how fast it is, and how Google and AI see it. Problems are sorted by what they cost you.

  • Works. Shops and signs up like a customer, on a phone and a desktop
  • Speed. What makes each page slow on a phone
  • SEO. Each page as Google and AI assistants read it
  • Suggested fix. For your platform: Shopify, WooCommerce, WordPress, Next.js, Webflow and more. Or we fix it for you

Speed, measured on a phone

  • Largest Contentful Paint
  • Layout shift
  • Server response
  • Render-blocking scripts
  • Heavy images
  • JavaScript weight

SEO, as a crawler sees it

  • Titles and descriptions
  • Headings
  • Canonicals
  • noindex by mistake
  • Sitemap and robots.txt
  • Structured data
  • Image alt text

AI visibility

Can AI assistants find your site?
Now you know.

More customers ask ChatGPT, Claude or Perplexity before Google. If their crawlers cannot get in, you are not in the answer.

  • Which crawlers your robots.txt lets in: Google, Bing, ChatGPT, Claude, Perplexity
  • Which pages show their content without JavaScript
  • Whether you have an llms.txt, and what belongs in it

The full audit

Tested like a customer.
Every step to the order.

The agent walks every path a customer can take, from the first page to the order or the sign-up, and finds what breaks before a customer does.

  • Signs up with its own inbox
  • Logs in, one-time codes and magic links too
  • Products, search, cart and discount codes
  • Pays with test cards, never a real charge
  • On a phone and a desktop
  • Every issue with steps, screenshots and logs

How it works

From an address to proof,
without touching your code.

  1. 01

    Enter what you run

    Your site, shop or web app address, your MCP or API endpoint, and a test account for pages behind the login. Nothing to install.

  2. 02

    The agent shops

    It uses your site or app like a first-time customer, on a phone and a desktop, and calls your API and MCP endpoints.

  3. 03

    You get proof, and a suggested fix

    A screenshot of every step and a suggested fix. Send it to your developer.

Track daily

Your site and app change every day.
Track them daily.

A website, shop, web app, MCP or API is never finished. Three reasons a check now and then is not enough:

  1. It changes without you. Theme updates, plugins, marketing scripts, releases, API changes: any of them can break checkout, sign-up or an endpoint.
  2. Nobody tells you. Customers do not report a broken checkout. They leave, and you see it weeks later as a dip in orders.
  3. Every day it stays broken costs orders. Tracked daily, you know the next day, with the screenshot and a suggested fix.
  • Each fix confirmed on the next run
  • Daily tracking uses no credits, on every plan

The difference

A speed score is not
the same as a sale.

Speed and SEO checkers, code-reading AI

  • Scan one page at a time, or read code and guess
  • List warnings, not what breaks the order
  • Never add to cart, sign up or try to pay
  • Report numbers and lines of code

WitnessQA

  • Uses the whole site like a customer
  • Finds the dead button and the failed discount code
  • Signs up and pays with test cards, on a phone and a desktop
  • Reports screenshots and steps anyone can follow

Pricing

Every plan starts with 7 days for $1.

Why $1? We use it to verify that you are a real user. Cancel within 7 days and you pay nothing more.

Starter

$19 / month

billed monthly

For a solo builder shipping fast

  • 500 credits a month, about 25 audits
  • Full audit, including flows behind sign‑in
  • Regression up to 10 times a day
  • 1 project, 1 person, 2 runs at once
  • 14 days of evidence
  • API and MCP for your own agents
Start 7-day trial for $1
Most popular

Pro

$49 / month

billed monthly

For teams that ship weekly

  • 1,500 credits a month, about 75 audits
  • Full audit, including flows behind sign‑in
  • Regression after every deploy, up to 50 a day
  • 3 projects, 3 people, 3 runs at once
  • 30 days of evidence
  • API and MCP for your own agents
Start 7-day trial for $1

Business

$99 / month

billed monthly

For products with many flows

  • 3,000 credits a month, about 150 audits
  • Full audit, including flows behind sign‑in
  • Regression after every deploy, up to 300 a day
  • Unlimited projects, unlimited people, 10 runs at once
  • 90 days of evidence
  • API and MCP for your own agents
  • Priority in the queue
Start 7-day trial for $1

7 days for $1 on every plan, then the price shown. Cancel anytime, no lock-in.

Why $1? We use it to verify that you are a real user.

Compare every limit, credit packs and the FAQ

FAQ

Questions, answered.

How is this different from PageSpeed Insights?

PageSpeed measures one page, once. WitnessQA uses your whole site like a customer, every day: it shops, signs up and pays with test cards, on a phone and a desktop, so you know the moment something breaks, before your customers do, with proof and a suggested fix for your platform.

How do you measure speed?

In a real browser on a phone with a slowed-down network, the same setting as Lighthouse and PageSpeed Insights. A page that looks slow is measured twice before it is reported.

What is a credit?

One step of the AI agent: open a page, click, type, check. A public audit uses about 20, an issue check 10 to 40. Speed, SEO, link checks and daily replays use none.

What does a full audit cover?

Every path a customer can take: product, search, cart, discount codes, checkout with test cards, sign-up, login, account and password reset, on a phone and a desktop. Plus links, speed and SEO on those pages.

Do you need access to my code?

No. A URL, a test account for pages behind login, and for issue checks, read access to your issues.

Do I need a developer?

No. The report says what to change in plain words, for your platform. If nobody is at hand, we fix it for you, paid per fix.

Will it pay for things or delete data?

Not on your live site. Destructive actions are blocked there by default, and no real card is ever charged. Checkout is tested with the official test cards of Stripe, Paddle and PayPal.

How does it log in?

The agent makes its own account with its own inbox, uses a test account you give it, or a TOTP secret for 2FA. One-time codes and magic links are read automatically.

Can I keep the tests?

Yes. Export every flow as Playwright code. No lock-in.

How does the 7-day trial work?

Pay $1 to verify your card and get the whole plan for 7 days, then the price shown. Cancel before then and you pay nothing more.

Can my own AI agent use WitnessQA?

Yes. Make an API key and add WitnessQA to Claude Code, Cursor or any MCP client, or call the HTTP API. Paid from the same credits.

How do you avoid false alarms?

Every finding rests on an objective signal, such as an error or a button that did nothing, and is reproduced on fresh runs. Only what shows up every time is an issue; the rest is marked worth a look.

Tomorrow,
know what got better.

Why $1? We use it to verify that you are a real user. Cancel within 7 days and you pay nothing more. Want to try first? Run a free audit of your public pages, no sign-up.