Ask why the number moved.Get the answer and the queries.
Point an agent at a read-only replica in a terminal card and your dashboard in a browser card. It writes the hypotheses on a todo card, checks each with SQL you can watch run, marks what it rejected — and leaves an answer in a note that anyone can re-run.
Four hypotheses, three queries, one cause. Allow the long query when it asks.
The agent checks the dashboard, turns your question into hypotheses, tests each one with SQL on a read-only replica in a terminal card, and writes the answer — with the queries — into a note.
Hypotheses you can see
Each idea is a line on a todo card — green when the data supports it, red when it doesn’t.
Queries in the open
Every query runs in a psql terminal card; point it at a read-only user and nothing can be written.
Checked against the dashboard
The agent opens the same dashboard you use in a browser card, so the answer starts from your number.
An answer that can be re-run
The note holds the finding, the table and the exact queries.
From a question to an answer
Four cards, one agent.
Works with what you already have
Agents for the rest of the work
The same cards run your inbox and your routines. Each page below is a live, clickable scenario.
Eight ways to ship with a squad
Each page is a live, clickable run of a real development workflow — with the same cards and tools the app gives your agents.
Give your next data question to an agent
It shows every query it runs.
macOS and Linux are coming later.
The first-run wizard installs an agent CLI, Git and the dictation model for you — or skips each step.