Stop paying for Azure resources nobody is using.
It runs inside your own subscription, shows you what it would change, and touches nothing until you let it.
78
kinds of waste in the catalogue
4
switched on until you choose more
0
credentials you hand over
It runs in your subscription
It runs in your own Azure account, using access you create there. Nothing to hand over, and no copy of your estate sitting in someone else’s cloud.
It will not touch production
Only what you tagged non-production. No tag means no touch, because a resource nobody labelled is one nobody vouched for.
It shows you what it refused
Every resource it left alone, and the rule that stopped it. Most tools only show you the wins.
It can be undone
While requireSnapshot is on, a disk is snapshotted before deletion and the delete is abandoned if that fails. Turn it off and the deletion is permanent, which the agent then labels as having no way back.
Another 74 checks ship disabled
The wider catalogue starts disabled, because a rule you did not ask for should not act on your estate. Turn on what you want once you trust it.
It writes down why, before it does anything
Two records from a real run. The first is an action it took. The second is the same check on the same machine, refusing.
Acted
rule idle_vm
resource vm-dev-idle
action deallocate_vm
evidence avgCpuPct 2.48, maxCpuPct 41.63
over 2h, threshold 10%
saving 8.22 USD / month (estimated)
rollback start_vm
result successRefused
rule idle_vm resource vm-dev-idle action deallocate_vm stopped by confidence_window reason observed 1/2 consecutive cycles result skipped
The refusal is the one worth reading. It had found the machine, priced the saving, and still would not touch it, because it had only seen it idle once. Every one of those records lands in a log on your own infrastructure.
What it will not do
Touch production
Or anything you have not tagged. 69 of the 78 checks only report, and never act at all.
Buy anything
Reservations and savings plans are the biggest lever in most Azure bills. That is a multi-year commitment and a person should make it.
Work outside Azure
No AWS, no GCP. If you need one view across all three you need a different tool, and you can run this underneath it.
Optimise Kubernetes
Cluster bin-packing and autoscaling is its own discipline with better tools for it.
Start with a dry run
One non-production subscription. It reports what it would have done and what it refused to touch, and calls no write API at all until you change that.
See what it would change