Pilot store · Discount codes
A discount code had been broken for over a week.
The code stopped working at checkout and nobody on the team noticed, because nothing else looked wrong. The owner guessed it had cost a couple thousand dollars in sales.
Free pilot, open to a few more stores
Find broken flows before your customers do.
Write the flow in one sentence, like a sign-up or a checkout. Clickwalker goes through it in a real browser every few hours and tells you in plain English when it stops working.
You don’t write or maintain test scripts. Runs on Claude.
Checkout flow1 issue found
- Open the product page
- Add “Canvas tote” to cart
- Go to checkout
- Fill in shipping address
- Enter test card 4242…
- Click “Pay now”: nothing happened
- Console: TypeError in checkout.js:212
Customers can’t pay right now. The “Pay now” button doesn’t respond. Screenshots and the console error were sent to #alerts.
- 15+
- real bugs caught by pilot runs
- 1 sentence
- is the whole test
- 0
- test scripts to write or fix
- Every few hours
- in a real browser, on a schedule
Write the test in one sentence.
If you can explain a flow to a new teammate, you can give it to Clickwalker.
-
1
Describe the flow
In plain English:
Sign up with a new email, then buy the cheapest item using the test card. -
2
Clickwalker runs it on a schedule
Claude drives a real browser. It reads each page, decides what to click and checks that each step actually worked.
-
3
You hear about breaks first
A report in Slack or email: which step failed, what the agent saw, screenshots and console errors.
Every alert comes with proof.
Passing runs stay quiet. When a flow breaks, you get one message with everything needed to fix it, written so anyone on the team can follow it.
- The step that failed, in plain words
- A screenshot of every step, so you can see what the agent saw
- Console errors and failed requests captured during the run
- A full report in Markdown and JSON to share or file as a ticket
ClickwalkerApp9:41 AM
Clickwalker: failed on yourstore.com
Buy the canvas tote. Ship to Test Customer, 12 Market St, Springfield. Pay with test card 4242 4242 4242 4242.
Customers can’t pay right now. On the payment page the “Pay now” button doesn’t respond, and no order is created.
uncaught TypeError: Cannot read properties of undefined (reading 'token') at checkout.js:212:41
- 01
- 02
- 03
- 04
- 05
- 06
Change the design and the test still runs.
Scripted tests
- Break when a button is renamed or moved
- Need an engineer to write and maintain them
- Fail with stack traces only developers can read
Clickwalker
- Finds the button the way a person would
- Anyone on the team can write a flow
- Tells you what went wrong in normal words
What the pilot has caught so far.
Clickwalker is in a small free pilot with under 15 stores, mostly on Shopify and WooCommerce. Runs have caught 15+ real issues. Two of them:
Pilot store · Checkout
A theme update dropped the shipping address.
After the update, the Stripe checkout stopped passing the address along, so orders came in with nowhere to send them. Clickwalker caught it by placing a test order end to end.
Unit tests passed in both cases, because each piece worked on its own. You only saw the bug by going through checkout like a customer.
For small teams without a QA engineer.
Online stores and SaaS products where a broken sign-up or checkout means lost money, and nobody finds out until a customer emails.
There’s room in the pilot for a few more. Send us your site and we’ll set up the first flow with you.
Built on Claude, with guardrails.
An agent clicking around your store should be predictable. Here’s exactly what it can and can’t do.
-
A real browser
Each run opens your site in a fresh Chromium session, with no saved logins or cookies from earlier runs.
-
Claude checks every step
After each action it gets a screenshot, a numbered list of what’s clickable and any new console errors, then decides what to do next.
-
Seven actions, nothing more
Click, type, press a key, scroll, wait, open a page and finish. The agent has no way to run its own code.
-
Stays on your domain
It can only navigate to addresses on the site under test, and it’s warned right away if a click ever leads somewhere else.
-
Test data only
It uses only the test accounts and test cards written in your flow, and treats text on your pages as data, never as instructions.
-
Fits in CI
Exit code 0 when a flow passes, 1 when it fails and 2 when it’s blocked, so a broken checkout can stop a deploy.
Simple pricing.
Free while we’re in the pilot. No card needed to start.
Pilot
$0
Open to a few more stores right now.
- We set up your first flow with you
- Runs every few hours in a real browser
- Alerts in Slack or email with screenshots
- Full report for every failed run
After the pilot
$29/month
Planned from Q1 2027.
- Everything in the pilot
- Flows written in plain English
- No test scripts to maintain
Questions.
Is it safe to run on my live store?
Clickwalker uses only the test data you put in the flow, like a test account and a test card such as 4242 4242 4242 4242. It never enters real personal or payment details, and it can only open pages on your own domain. You can point it at your live site with payments in test mode, or at a staging copy.
Do I need to write any code?
No. A flow is one sentence in plain English. During the pilot we set up the first flow with you.
Which platforms does it work with?
Anything that runs in a browser. Clickwalker doesn’t plug into your platform; it uses the site the way a customer does. Most pilot stores are on Shopify and WooCommerce.
What about cookie banners and popups?
It closes them the way a person would. Pages that change a lot between runs, like A/B tests and app popups, are the hardest case, and it’s what we’re improving most right now.
What if Clickwalker gets it wrong?
It can. An AI agent can miss a bug or flag one that isn’t there. That’s why every alert includes a screenshot of each step, so you can check what it saw in a few seconds. Treat it as an extra safety net next to your own monitoring.
What happens to my data?
Screenshots and page text from runs go to Anthropic’s API, which runs the Claude model behind the agent. We keep run data only as long as we need it for your reports and delete it whenever you ask. Details are on the privacy page.
Which flow can’t break?
Send us your site and the one flow you’d hate to see broken. We’ll set it up with you.