​

Crawling and the Pages catalog

The Pages catalog is Qodex’s map of your web app. It shows which pages Qodex has discovered, what each page looks like across screen sizes, and which backend endpoints the page calls while loading. This gives the agent real product context before it writes UI tests. Instead of guessing which routes exist, Qodex can ground scenario creation in pages it has actually visited.

​

How crawling works

Qodex starts from a URL you provide and walks through same-origin pages in a deterministic breadth-first crawl. For each page, it captures the page title, responsive screenshots, links for the next crawl step, and backend requests fired by the browser. The crawler finds pages from three signals:

• Anchor links that point to other pages in the same app.
• Click discovery for buttons and menu items that change routes without normal links.
• Scenario runs that visit a page the crawler has not seen before.

The crawl uses the agent’s current logged-in browser session, so authenticated pages can be discovered the same way a real user would see them.

​

What Qodex stores

Every discovered page becomes a row in the Pages catalog.

​

Responsive screenshots

Qodex captures each page at three viewport sizes: These screenshots help you understand what Qodex saw during discovery and can reveal responsive issues before you even create a scenario.

​

API calls discovered from UI pages

While the crawler visits pages, it also watches network requests. Backend calls are deduplicated and added to the endpoint catalog, so UI discovery can improve API coverage too. This is useful when your OpenAPI spec is incomplete or when frontend teams ship routes before the API catalog is updated.

​

When this matters

• You are onboarding a new app into Qodex.
• You want to see which pages have test coverage and which pages are still untested.
• You want the agent to create scenarios from pages that actually exist.
• You want screenshots across screen sizes without setting up a separate visual pass.
• You want observed API calls to feed the endpoint catalog automatically.

​

Related