What is Aegis?

The five-minute version, in plain English. No code, no acronyms. If you need the detail afterwards it is all in User Guides and Core Concepts — this page is the thing you send somebody first.

---

1. In one paragraph

Aegis is the layer between an ad click and a landing page.

It sits at DNS level. The subdomain in the ad — meta.purdyandfigg.com, youtube.purdyandfigg.com — points straight at Aegis, so the request reaches us before anything else. We work out where that page really lives, serve it under our own domain, attach the tracking it needs and record the click as ours — then keep checking, every hour, that the page still works.

Nobody is redirected anywhere. The customer never leaves the brand domain.

---

2. The page is ours, wherever it was built

Marketing builds where it likes. A page might be a Shopify page, a React app built in Lovable, a static site on Cloudflare Pages, or one Aegis serves itself — and the visitor never finds out which. They see one domain, ours, and the tracking behaves identically on all of them.

And a page can move. Something built quickly in Lovable can later be promoted — the same route, the same public URL, served from Shopify or Cloudflare instead. The tracking, the tests and the monitoring all carry over, because none of them were ever tied to where it happened to be built.

That matters for two reasons, and only one of them is tidiness:

• The click belongs to us. It is recorded first-party, on our own subdomains, rather than borrowed back from a platform that has its own reasons to round up.
• Nothing is stranded. A page built in a tool we stop using is still a page we serve, still tracked the same way, still tested the same way.

---

3. It works with Shopify — and can sell without it

Shopify stays the backend of record: stock, fulfilment and customers never leave it. What Aegis changes is where a page is built, and where the sale can happen.

Linking to Shopify's checkout. Cart permalinks are fiddly to hand-write, so they are built from a screen instead — variants, quantities, selling plans, discount codes and UTMs. Every built link is stored, which is what lets the hourly checkout test prove it still opens a real cart at the right price, and each one gets a tidy shortlink (/checkout/summer-kit) rather than two hundred characters nobody can read or verify.

Or selling on our own page. The Ecommerce module gives a landing page its own cart and its own checkout, with Shopify out of the path entirely — an Aegis-hosted page. Each product is matched to its Shopify variant, so a sale still moves stock and still shows up in Shopify's product reporting, and the moment payment succeeds the order is written into Shopify for fulfilment.

One rule underneath all of it: prices come from the database, never from the page. The page says which product and how many; the server says what it costs. A storefront that accepts a price from the browser can be bought from for nothing, and the order looks completely normal afterwards.

---

4. It watches itself

A live page is checked constantly, and every check is a switch you can turn on per page rather than a setting for the whole estate:

Is it up?	Repeat failures escalate, rather than pinging once and giving up
Is the tracking there?	Pixels and tracking attributes, on the page as served
Does it still look right?	Today's screenshot against a saved one
Does the buy button work?	The link resolves to a real cart, right product, right price
Is it too heavy?	Oversized images flagged
Is the ad copy compliant?	A Passmark scan, on the pages where it matters

Cheap checks run every hour. The expensive one — an actual test of the page — runs on its own schedule underneath.

---

5. The testing is done by an AI agent with a real browser

Most of our landing pages are built by JavaScript running in the browser, so an ordinary automated test fetches an empty shell and declares victory. Ours drives a real Chrome, on a real machine, and sees what a customer sees.

Three things follow from that, and they are the parts worth knowing:

• A test is written in plain English, and lives in the database rather than in code. Anyone can read one, run one, or change one. No developer required.
• A person always signs it off. The agent may write a test, run it, and say what it thinks is wrong. It cannot make its own test the live standard, and it cannot close its own findings. That line is deliberate and it does not move.
• When the agent gets something wrong, the fix is a paragraph, not a code change — a shared playbook every test run follows. And a rule now needs a second person to approve it, because it reaches every page at once.

---

6. What happens when something is wrong

A finding is a thing with a life, not an hourly alarm.

When a page fails, somebody decides which of two things happened — the page is broken or the test is out of date — and that decision is recorded rather than remembered. Getting it backwards is the one genuinely dangerous mistake in QA: updating a test to match a broken page makes the alarm stop and hides the fault.

From there:

• A real problem becomes a defect, with a severity and an owner, tracked until somebody fixes it.
• Seeing it again counts, rather than filing it again. One problem stays one row with a number beside it.
• Deciding to live with something sticks. It cannot quietly re-open an hour later.
• If a later check can't reproduce it, that is recorded on the defect too — so whoever picks it up knows the last look disagreed with the filing.

Everything open across both storefronts lands on one shareable page, grouped by store, then severity, then who needs to fix it. It says at the top how much of the estate has actually been checked, because a short list means much less when half the pages have not been looked at.

---

Where to go next

You want	Read
To run QA day to day	QA Manager Handbook
To write or review a test	QA User Guide
To add or change a page	Admin User Guide
A specific question	Help & FAQ
Who Agatha is	Meet Agatha

Aegis runs on eight small services sharing one database, serving two storefronts, UK and US.