Got tired of setting up dashboards just to answer "where does our K8s money go?"
So I built Burn — a CLI that shows you exactly which namespaces,
pods, and load balancers cost what. One command, no agent, no dashboard.
## What it looks like
$ burn analyze --prometheus http://prometheus:9090
NAMESPACE PODS CPU REQ→USED COST/MO
argocd 4 2.0 → 30m $56
amazon-cloudwatch 11 1.6 → 82m $44
kube-system 21 1.4 → 52m $41
...and 7 more namespaces
Idle (unallocated) $117
Total $350
COST BREAKDOWN
Compute: $350 | Storage: $0 | LB: $16
Total: $367
argocd-dex-server requests 500m CPU, uses 0.12m. That's $14/month for 0.02% efficiency.
## Slack integration
It also works as a Slack bot. You just ask:
/burn ask "what should I do to save money"
→ 1. Switch to Spot — Save $277/mo
2. Rightsize argocd-dex-server (500m → 10m CPU)
$ kubectl set resources deployment argocd-dex-server -n argocd --requests=cpu=10m
3. Delete forgotten rds-debug pods — $11/mo wasted
You ask a question, it gives you kubectl commands. No manual analysis.
## How it works
- Pulls node and pod data from the Kubernetes API
- Gets actual CPU/memory usage from Prometheus
- Fetches real pricing from AWS and Azure APIs
- Splits cost per resource (CPU and RAM priced independently)
- Detects load balancers via both Service and Ingress resources
- GPU nodes detected and priced automatically
## Install
brew install tanrikuluozlem/burn/burn
Also available as Docker image, Helm chart, or Go binary.
## On-prem and GPU
Works with on-premise and GPU clusters too:
burn analyze --cpu-price 0.05 --ram-price 0.008 --gpu-price 3.00
Open source, Apache 2.0: github.com/tanrikuluozlem/burn
What does your cluster idle cost look like? Curious if 33% waste is normal.