Phrontis

Know what every namespace costs.

Kubernetes cost per namespace and workload, idle capacity, and the requests to change.

Free for 30 days. Read-only agent.

See where the bill goes

Cost per namespace and workload, and how much node capacity sits idle. AWS and Azure nodes are priced from the vendor’s public list price.

Phrontis overview of a demo cluster: $503.70 a month, $145.26 of it idle, 71.2% of capacity allocated, and the top recommendations.

Right-size in a dry-run first

Each request is set to peak usage plus 20% for CPU and 30% for memory. A dry-run shows every change and what it saves before anything is applied.

Phrontis autoscaling dry-run: each container’s request before and after, and $151.32 a month the change would save.

Every change in a journal

Each change is logged with the spec before and after. The kill switch stops automation and leaves the cluster as it is.

Phrontis journal: changes listed as would apply, recorded and applied, with the spec before and after.

How the pilot works

Free for 30 days on one of your clusters. You get every feature, automation included.

  1. 1

    Request a pilot

    Leave your email, company and cluster size.

  2. 2

    Install the agent

    One helm install. It only reads.

  3. 3

    Read the report

    Cost by namespace and workload, idle capacity and requests to change.

  4. 4

    Try automation

    Workload and node autoscaling, each in a dry-run first.

Cost report
Cost per namespace and workload, idle capacity, and spot candidates.
Rightsizing
Requests to change, with OOM and throttling guards.
Node policy
On AWS, a ready Karpenter NodePool for the cheapest fleet.
Automation
Workload and node autoscalers, the journal and the kill switch.

Pricing

After the pilot: Free or Pro.

Free

$0

Report mode, up to 500 vCPU.

Pro

$2

Per vCPU per month.

Questions

What can the agent read?

Its ClusterRole has only get, list and watch. It reads nodes, pods, volumes, services, workloads and pod metrics, and no Secrets, ConfigMaps or logs. It sends its samples only to the API address you give it.

Will it change my cluster?

Not until you turn automation on. Every automation starts as a dry-run. The node policy is applied only after Phrontis shows the diff and you confirm, inside your maintenance window.

How do I install it?

Sign in with GitHub, then run one helm install. The agent image is signed with cosign.

Are the costs my cloud invoice?

No. They are estimates from public list prices in USD, before discounts or savings plans.

What if a container runs out of memory?

A container whose last exit was an OOM kill never gets a memory request below its current request. A container whose CPU peak reaches 90% of its limit is flagged as throttled.

See your cluster’s costs in a free 30-day pilot.

Start a free pilot

30 days on your cluster, free. We reply by email with the next step.