What you can ask

Ask in your own words. Your AI tool picks the Kubeadapt tools it needs, often several in a row, and answers with the figures, the time window and how they were measured. Naming the cluster, namespace or team and the period you mean gives the most precise answer.

Where the money goes

• Which of my clusters cost the most this month, and how does that compare with last month?
• Break down spend on the prod cluster by namespace for the last 30 days.
• What are the ten most expensive workloads across all clusters, and what does each cost per month at today's rate?
• How much do we spend on storage, and which volumes cost the most?
• What does the team=payments label cost us? List the label values we use.
• Where will this month's spend end up if nothing changes?

Why it changed

• Why did last week cost more than the week before? Which workloads drove the change?
• Our bill jumped on the 14th. What changed that day?
• Compare September with August by namespace.
• Is anything unusual in this week's spend for the data-platform namespace?

Teams and chargeback

• What did each team cost this month, and what share is still unassigned?
• Which workloads have no team yet, starting with the most expensive?
• Give me the platform team's month so far, its top workloads and its projected month-end.

Waste and idle capacity

• How much of our bill is idle capacity that no workload requested?
• Which nodes are the least utilized, and what do they cost?
• What would a smaller node layout save on the prod cluster?
• What would we save by moving half of our on-demand capacity to spot?
• Which volumes are not attached to anything but still cost money?

Rightsizing

• Which workloads are the most over-provisioned, and how much would rightsizing save?
• Get the rightsizing plan for the api deployment in the shop namespace and give me the YAML patch.
• Show me CPU and memory use of redis over the last two weeks against its requests and limits. When does it peak?
• Which containers were OOM-killed or restarted, and what should their memory be?

Alerts and budgets

• Are any incidents open right now? What fired, and is the cost change real?
• Which budgets are we close to, and how much of each is spent?
• What alert rules do we have, and which ones fired in the last 30 days?

Network cost

• What does cross-zone traffic cost us, and which workload pairs cause most of it?
• Which internet destinations do we send the most data to?
• Who does the checkout service talk to, and what does that traffic cost?

Pricing and accuracy

• Why does this cluster cost what it does? Which rates is it priced with?
• Are any of our clusters priced with Kubeadapt's default rates instead of our real prices?
• What is the carbon footprint of each cluster this quarter?

Ready-made prompts

Kubeadapt MCP includes prompts that run a complete review in one step. Most AI tools list them in their prompt or slash-command menu.

Tips

• Say the period. "This month", "last week" and "September" all work. Without one, most tools use the last 30 complete days in UTC.
• Ask for the cost mode you need. Fully loaded cost, the default, includes each workload's share of idle capacity; workload-only cost is what the workload itself uses or requests. See Cost modes.
• Ask it to open Kubeadapt. Every answer includes a link to the same view in the app.
• Read the caveats. If a cluster stopped reporting, a price is missing or a cluster sends no network data, the answer says so instead of showing zero.