Cost control · Cost Spike Sentinel
Native anomaly detection is subscription-wide and capped at five rules. This gives every resource group or tag value its own baseline and its own alert.
Deploys as a managed application in your own subscription. No vendor backend, no data egress. Flat monthly fee, billed by Microsoft.
A four-times jump in one team's resource group is invisible in a large, steady subscription total, and you cannot stand up an anomaly rule per team because Azure caps you at five. The people who could fix a spike are the last to hear about it.
A serverless function on a schedule, with a system-assigned managed identity, deployed into your subscription by the Azure Marketplace. Roles are granted by you after deployment with the one-line script in the package, so nothing runs until you say so. Every role it asks for, and why:
| Cost Management Reader | read daily cost per slice |
| Reader | resolve resource groups and tags |
Alerts go to the Teams webhook you configure. Nothing is sent anywhere else.
No. The function runs in your subscription and writes only to your configured alert destination. Katabarwa Labs operates no backend and receives nothing.
A flat monthly fee on your Azure bill through the Marketplace, plus the small consumption cost of the function itself. No per-resource or per-user pricing.
Delete the managed application from the Azure portal. Its resource group and identity go with it, and any role you granted can be removed the same way.
support@llmgraph.ai, or the support page.
Native anomaly detection is subscription-wide and capped at five rules. This gives every resource group or tag value its own baseline and its own alert.
Deploy from the Azure Marketplace