Spend over time
By tool
Top spenders
| User | Month |
|---|
| User | Month |
|---|
How often you're reminded once your AI spend passes 75% of your tier cap. The in-app bell and via Slack. A tier change always sends a fresh alert.
The organization's monthly spend limit per provider. The Everyone page's By tool pane uses it as that provider's budget: the bar fills toward the limit and its shade darkens once usage passes 75% of it. Only Claude is editable for now.
Each tier's monthly Claude cap. A change takes effect right away for everyone on that tier. It updates their My Tier limit, the 75% alert, and any webhook. Tier 4 is a custom per-person limit (see View Tier 4 Limits).
Tier 4 is a custom per-person limit. Assign each person their monthly Claude cap. Saving records it here; the limit must also be changed in Claude by hand.
Set any person's Claude spend tier directly — up or down. This moves their Entra tier-group membership through the tier automation and messages them the new tier. Distinct from a tier request, which only bumps someone one tier on an approval.
View the app as another user (their usage, tier, and tabs) to see exactly what they see. A red bar stays at the top while active. Admin + dev only; it does not exist in production.
The default cadence for the 75% over-limit alert. Each person can override it for themselves in their own Notifications settings.
Fired once per person when their AI spend crosses 75% of their tier cap. The same trigger as the in-app bell. Leave disabled to send nothing.
Fired when someone files a tier-increase request, to notify their manager. Leave disabled to send nothing.
Fired when an admin sets someone's tier to any value from the Tiers table. Points at the JSD-Set-User-Tier Logic App, which moves their Entra group membership to the target tier. Leave disabled to hide the Set control on the Tiers table.
Pull usage from the providers now instead of waiting for the hourly job. Leave the date blank for the default window, or set it to backfill from a start date. Runs in the background; refresh in a minute to see the data land.
Codex reports token usage but no dollar cost (OpenAI seats are flat-rate), so its spend is simulated from these per-model rates. Figures are cents per 1,000,000 tokens — e.g. 200 = $2.00 / 1M. A model not listed simulates at $0.
Click a row to see the approver, decision and justification.