What placemy actually does.
Everything on this page ships in the binary today. We measured 24–67% savings across five real workloads in the dissertation evaluation. Here is how.
Multi-cloud VM discovery
Finds every running EC2 instance, Azure VM, and GCP Compute Engine instance across all accounts you authenticate. No agents required — uses your existing cloud credentials.
CPU and memory utilisation profiling
Pulls real metrics from CloudWatch, Azure Monitor, and Cloud Monitoring. Calculates average, P50, P95, P99, and peak utilisation. Not a guess — measured data from the last 24 hours (or up to 14 days).
Works with what you already have
Uses the standard credential chain (AWS SSO, az login, gcloud auth). If your terminal can talk to your cloud, placemy can scan it. No separate service account, no IAM role to build.
18,000+ instance types compared
Pricing catalogues from AWS and Azure, refreshed daily. Every candidate is filtered by OS, region availability, and current-generation status. No manual spreadsheet work.
P95-based sizing, not peak or average
Average CPU of 5% and peak of 90% could mean either a bursty workload or a steady one with a single spike. We use P95 to size correctly for the workload's real pattern, not the outlier.
Memory-aware recommendations
If you have a monitoring agent installed (CloudWatch Agent, Azure Monitor Agent, Ops Agent), we factor in memory pressure. We never recommend downsizing memory on a VM that's using 80%+ of it.
Cross-cloud comparison
The same workload might cost 67% less on a different provider in the same region. We check both directions — AWS to Azure and Azure to AWS — and show you the numbers.
Self-contained HTML report
One file, no internet required. Open it in a browser and see every VM, its utilisation profile, the recommended alternatives ranked by savings, and the evidence trail explaining why each is safe.
Machine-readable JSON
Every recommendation is available as structured JSON for your own automation — pipe it into Terraform, feed it to a FinOps tool, or build a Slack notification.
Completely read-only
Scanning never mutates anything in your cloud account. No resources are changed, no tags are written, no instances are stopped. You decide what to act on after reading the report.
Storage optimisation (S3/Blob lifecycle, orphan EBS volumes), reserved-instance and savings-plan modelling, and container workload support are in development. They are not on this page because they are not in the binary yet.