Shared sources of truth
Warm Queries
Goal: Crystallize the work that repeats into a shared source of truth.
A warm query is a named, saved query against your data. This is where repeatable work gets crystallized: one teammate (or one AI) writes a good query, saves it, and it becomes the answer everyone's AI builds from.
WHY IT MATTERS
One query, three things at once
Saving a warm query automatically makes it an API endpoint and an MCP tool, and records who approved it. One good answer becomes a shared source of truth your whole team - and any AI you connect - can use.
Build it
Inside a surface, drag a Warm Query node, wire data to it, open the Warm Query Manager, and write the query yourself. Run a preview, then save and deploy. New to the query language? See KQL Basics.
From any MCP agent: it tests query ideas with run_adhoc_warm_query, then proposes the saved
query via manage_proposals. You accept it in the portal and deploy.
In the Warm Query Manager, ask the scoped Azi: "average signups per hour over the last hour." She proposes it; Run to preview, approve, deploy. More on phrasing asks -> Working with Azi.
The query runs when called - warm and ready, not constantly.
Read it back
Once deployed, your warm query is reachable three ways: preview in the surface; REST
(GET /oi-api/warm-queries/{lookup}); or MCP (run_warm_query from any AI). The
catalog of your queries is the warm_query_catalog resource.
Build the next from it
Because any AI can both read (run_warm_query) and propose (manage_proposals), one source of
truth becomes the seed for the next. Accept proposals in the portal and the workshop compounds.
Every decision is remembered
Your request, the proposal, your approval, and the timestamp are recorded - who decided what, and why. See Ship with Confidence.
Next steps
- Learn the query language -> KQL Basics
- Build queries with Azi -> Working with Azi
- Call it from anywhere -> APIs
- Connect any AI -> Bring Any AI