How it works

Five layers, each derived from the one above it

Each layer is generated from the one above it. Owners correct the layer where they have the knowledge, and the correction flows back up. Tests are the last layer, not the first.

Tests are the last layer, not the first. Each layer is generated from the one above; corrections from the owners flow back up, so the map gets better instead of the scripts getting patched.

The map

Dependencies are explicit, so a change already knows what it touches

The map · hover or tap a feature

changeddragged alongprerequisite

This is the thing nobody keeps today. Every test in the product hangs off one of these nodes; change a node and the map already knows which scenarios to re-check.

Architecture

Three engines, one feedback loop

Each engine bootstraps its layer once, then lives on feedback. Corrections sync upward, changes cascade downward, and the test runtime underneath is a pluggable choice.

Same colours as the whiteboard

engineinputmodule

Each engine bootstraps its layer once, then lives on feedback. Corrections sync upward; changes cascade downward; the runtime underneath is a pluggable choice.

Bring your own runner

Playwright, Selenium, Cucumber or an AI with a browser — the scenario tree is the product; the tests are outputs.

Get early access See how it works