stepscale is a self-hosted Kubernetes operator. The full analysis is Community and runs recommend-only: a rule engine plus LLM reasoning on your own key reads your HPAs and KEDA ScaledObjects and writes recommendations you review with kubectl. Team lets the operator apply the changes it recommends, behind a rollback safety net.
$ helm install stepscale oci://ghcr.io/stepscale/charts/stepscale-autoscaler \
--version 0.4.0 -n stepscale --create-namespace \
--set llm.provider=openai --set llm.apiKey=<key>
Recommendations appear about 30 minutes after metrics are flowing, viewable with
kubectl get scalerec.
No login, no license. Use
--set llm.provider=none
for rule-based findings without an LLM key. Full guide at
docs.stepscale.io.
No license, no signup, no credit card. Recommend-only analysis with full LLM reasoning is included at no cost.
When you trust the recommendations and want the operator to act on them.
Adds
Large estates, compliance requirements, air-gapped environments.
Adds