Use Origin safely and accurately from your AI assistant
Teaches an AI assistant to confirm the tenant and time window, read the schema before querying, and check truncation and cost before reporting numbers. It is read-only and never changes Origin data.
Use it when
- You want to know model spend or token usage by team over a set period.
- A query returned few rows or an error and you need it explained.
- You are checking which agents or endpoints were active this week.
Tools it uses
analytics_queryanalytics_describe_schema
What did we spend on each model over the last 7 days, highest first?
Over the last 7 days, your highest-spend model was claude-sonnet with 412 thousand tokens across 38 sessions, followed by gpt-4o with 150 thousand across 12 sessions. The result was complete, not truncated, and the window was set to 7 days explicitly. I did not include the sampled estimate for the smaller models.