PYX data product agentic engineering
Product tour

The real screens, in order

Captured from the running product, not mockups. For the clickable versions, the demos recreate four of these surfaces with scripted data.

Start from a sentence
Start from a sentence · The front door is a question in your own words. Or a spec, a report or a Power BI file dropped anywhere on the page. Nothing gets created until you say so.
Build from requirements
Build from requirements · Every row carries its journey as dots, and states its blocker in words rather than a status code: declared but not mapped, mapped but not governed.
The co-modelling canvas
The co-modelling canvas · Tell the model what it is missing, or drop in a workbook. Anything the agent proposes arrives as a dashed ghost you accept or dismiss, and nothing changes until you do.
Concepts, and where each one came from
Concepts, and where each one came from · A binding is the only place that names a schema, a table or a column. Everything else is stated over concepts, and every concept carries its origin: a workbook, a requirement, or declared by hand.
A governance workspace
A governance workspace · Nine controls bound, seven of them gates that hold the release, split across data quality, reconciliation, governance and lineage, and deployed as a job.
The whole estate as one graph
The whole estate as one graph · 106 nodes: requirements, concepts, models, tables, measures and platform connections, with approved, draft, proposed and unbound kept visibly apart.