The Workload Autoscaler has been the most valuable feature for us. Our Kubernetes workloads were heavily over-requested, and it right-sizes CPU and memory requests automatically based on actual usage, so we didn't have to tune hundreds of manifests by hand. That cut our requested capacity and node spend noticeably.

We also get detailed visibility into the cluster: usage vs. requests per workload, efficiency, and cost breakdowns. It covers everything we used Kubecost for, so we were able to drop it and keep one tool for both optimization and cost insight. There's also an MCP server, so we can query cluster and workload data straight from AI assistants without going through the UI or writing API scripts. Review collected by and hosted on G2.com.

The same data shows up in several dashboards, and the numbers don't always match from one page to another. That makes it hard to know which figure to trust or report internally, and we've had to cross-check against our own billing data. Some metrics are also hard to read without documentation, for example whether a value is per node or per template. Review collected by and hosted on G2.com.