Step 3: Build a Warm Query
Goal: Crystallize a warm query - your first shared source of truth.
Step 3 of 5. The crystallize moment. A warm query is a named, saved query that automatically becomes an API endpoint and an MCP tool. One teammate's good answer becomes the answer everyone's AI builds from.
Go inside the surface
- Click your
GrowthMetricssurface, then the Manage Surface icon (layered squares) - You are now inside the surface - the breadcrumb shows
{workspace} / GrowthMetrics
Add and wire the warm query
- Drag a Warm Query node onto the surface
- Draw a line from your data to the warm query so it has something to read
Build it with Azi
Open the Warm Query Manager (click the warm query). Its Azi panel is scoped to this query. Ask for what you want:
"Average signups per hour over the last hour"
Azi proposes the query and explains it. Click Run to preview real results, then Approve (or edit). Save and Deploy. Deploy takes a moment - your query goes live and API- and MCP-reachable on its own; continue to Step 4.
Azi: Here's
signups-by-hour- it averages signups in 5-minute buckets over the last hour. Run the preview, and approve when it looks right.
What you built
Data -> Surface -> Warm Query -> API + MCP tool. That is a shared source of truth. Next, read it back from any AI you like.