Plain-language questions, answered from the same 150+ Metric definitions every Report uses — so the figure in the answer is the figure on the page. And every answer opens the Report behind it, already filtered.
In the app when you're working, or from the AI you already have open.
A question box on every Report. Ask about what’s on screen or about anything else, and get the number with the reason behind it.
One address into Claude, ChatGPT, Copilot or Cursor. Your AI calls Kompound directly and gets the answer computed against live data.
Other tools hand your AI an export that’s hours old. The Connector runs the query at the moment you ask, on the same engine the Reports use.
The boundaries are the point. An analytics assistant that could write to your Store would be a liability.
Kompass cannot change a Cost, a filter, a Goal or anything in your Store. It answers questions.
No Customer names, emails, phones or addresses — not in an answer, not through the Connector, not in an export.
Kompass sees exactly the Stores your login sees. A Viewer on one Store gets answers about that Store and nothing else.
500 queries a month on Pro, 2,000 on Growth, 10,000 on Scale, under fair use. You see your usage and get a warning before the cap.
If a Cost isn’t set or a range has no data, the answer says so rather than filling the gap.
The link back to the filtered Report is not decoration. It is how you verify the number without trusting the sentence.
No. Kompass queries the same Metric registry the Reports are built from, so it cannot produce a number the Reports disagree with.
An MCP endpoint. You authorise it once inside Claude, ChatGPT or Copilot, and from then on your AI can query your Kompound data directly, read-only.
No. Queries are answered and discarded. Nothing from your Store is used for training.
You get a warning before it, and it's a soft cap under fair use — we'll talk before anything stops working.
Ask Kompass in the app, or ask from Claude and ChatGPT through Kompound MCP. Every answer opens its Report.