New Design Review: the cost of a Terraform change, on the pull request

Every dollar, traced to the team that spent it.

Cloud, Kubernetes, bare metal and AI coding agents flow into one ledger. CostOptix splits it by team, sends each spike and saving to the team that owns it, and prices infrastructure changes before they merge.

AWSAzureGoogle CloudKubernetesHostsCoding agents
One ledger · sample data $30,781.82
Platform team$12,480.35
Data team$8,240.52
ML team$6,910.18
Web team$3,150.77
Read-onlyNo write permissions, ever
AES-256-GCMCredentials encrypted at rest
~2 minutesTo your first dashboard
Flat rateNever a % of your spend

Questions in plain English.
Answers with the numbers behind them.

The questions finance and engineering ask every month, answered in seconds from your own spend. Pick one.

AI cost assistant Example answer
Why did our bill jump yesterday?

Built for the moment someone asks
“why is this so expensive?”

What you see on day one, shown with sample data.

#cloud-costs
CostOptix 09:02
Cost spike on Azure DevOps
$30.03 yesterday vs $11.38 expected +164%
HighAzureprod-subscription

Anomalies, the day they happen

Every service, namespace and host unit is checked against its own history. Alerts go to Slack, Teams, Discord or any webhook, and to the team that owns the spend.

Data teamSeptember
Kubernetes$5,120.40
Hosts$1,940.12
Coding agents$1,180.00
Total$8,240.52
Reconciled to invoice

One bill per team

Shared cluster and machine costs split by what each workload used, then reconciled to the invoice at month close.

Engineering$20,000 / month
$15,640 spent
78% usedwarn at 80%
Projected to reach the limit on Oct 24

Budgets that warn early

For accounts, tags, teams, clusters and hosts, with the day you are projected to cross the line. Sources without a budget get one proposed, ready to accept.

web-01today
postgres
38%
api
27%
nginx
9%
idle
26%

Hosts, split by what runs on them

VMs and bare metal broken down by service and process group, with idle capacity shown on its own.

Rightsizingaks-prod
Resize node pool
3 × D8ds_v5→3 × D4ds_v5
Estimated saving$412 / mo

Savings, ranked

Rightsizing, idle capacity, orphaned resources and commitment gaps on AWS, Azure and Google Cloud, ordered by what they save.

Coding agentsthis month
$1,180spend
14people
3unused seats
SonnetOpusHaiku

AI agent spend, per person

Claude Code and Gemini CLI usage by person, model and repository, valued at current list prices, with seats that go unused and tips when long sessions drive the bill.

All spendSeptember
Cloud bills$24,310.60
of which clusters$5,120.40
Hosts outside cloud$1,940.12
Coding agents$1,180.00
Total$27,430.72

Every source, counted once

Cloud, clusters, hosts and AI agents in one total. A cluster running in a connected cloud account is shown inside that bill, never added twice.

Checkoutcost per order
$0.042 per order
Down 28% since Maytarget $0.050

Cost per order, not just cost

Divide a team's spend, or any slice of it, by orders, customers or requests. Revenue can come straight from Stripe.

#team-data
CostOptix Mon 08:00
Weekly cost report: Data
  • etl namespace +$212
  • Engineering budget: limit by Oct 24
  • Held: disk cleanup to Jan 20

Reports that come to you

A weekly report for each team with what moved, budgets at risk, savings worth acting on and what was decided. A monthly summary for leadership.

Every cost has an owner. Every saving has a decision.

Teams claim spend by namespace, host, cloud tag, service, resource group or account. Shared costs like observability are split by use and managed by one team, so nothing sits unowned.

  • Spikes and savings go to the owning team's channel
  • Name the team that decides on clusters, commitments and AI plans
  • Hold or defer a saving with a reason and a revisit date. It reopens if the saving moves.
How ownership works
Open savingssample data
Rightsizing Resize node pool on aks-prod $412/mo
Platform (decides about this cluster) Open
Orphaned Unattached disk data-restore-01 $38/mo
Data Held until Jan 20
“Needed for the restore test”
Commitment 1-year compute savings plan $1,210/mo
Platform (decides about commitments) Deferred to Dec 1

Know the cost before you merge.

Add one step to GitHub Actions, GitLab CI or Azure Pipelines. Each Terraform plan is priced against what the resources really cost you today, not a list price, and the result lands on the pull request.

  • VMs, scale sets and node pools on AWS, Azure and Google Cloud
  • Try scenarios on a system: resize a pool, change replicas, remove a resource
  • Record the decision, then check the real bill against the estimate
Set it up in your pipeline
costoptix-botcommented on #482
CostOptix design review
+$153.56 per month ($466.92 now, $620.48 after apply)
ChangeResourceNowAfterPer month
Changeaws_eks_node_group.app3 x m6i.xlarge, $412.804 x m6i.xlarge, $550.40 †+$137.60
Addaws_instance.worker-m6i.large, $70.08+$70.08
Removeaws_instance.legacyt3.large, $54.12--$54.12

3 of 4 changes priced. Not priced: aws_db_instance (1). Resources that already run use what you really pay. New ones use list price.

† Scaled from what you pay today by the list price change.

See what every app costs, as it runs.

The host agent maps which apps call which, with each one's daily cost. Select one to see what would lose a dependency if it stopped, and how much daily spend that carries.

Kubernetes maps are in beta.

Take the tour.

Or click through it yourself in the live demo

Connected in minutes, not a project.

Read-only credentials for cloud, one command each for clusters, hosts and coding agents.

Cloud accounts
AWS    read-only IAM policy for Cost Explorer and billing
Azure  Reader and Cost Management Reader roles
GCP    Billing Account Viewer
  • Encrypted with AES-256-GCM
  • We never write to your accounts
  • Billing history imported on connect

Plans that don't grow with your bill.

Start free. Scale when ready. No percentage of cloud spend, ever.

Monthly Annual Save ~17%

Common questions

Find out who spent what.
Today, not at month end.

Free plan, no credit card. Connect a cloud account, a cluster, a machine or a coding agent and see your first numbers in minutes.