The bill went up, and nobody can say why.
CloudJaeger compares the periods and breaks the change into the services that moved. The rows add up to the headline, to the cent.
Architecture-aware cloud intelligence
See the cost.
Understand the system.
CloudJaeger explains what changed in your cloud bill — which service moved,
who owns it, and which fixes are worth your engineers’ time.
Frontier / Oak Ridge National Laboratory · Film creditsScroll to explore
Read-only connections
- AWSLive
- Microsoft AzurePreview
- Google CloudPreview
AWS runs on real bills today, through a read-only role. Azure and Google Cloud sync real cost and inventory, and stay in Preview until checked against a real bill.
All statusesThe problems we solve
A cloud bill says
what you spent.
Not why, or what to do next.
CloudJaeger turns the bill into answers your finance and engineering teams can act on.
Spend that nobody owns.
Costs are assigned to teams and products from your tags. Untagged spend stays visible as its own line instead of being spread across everyone.
Savings advice nobody trusts.
Each finding comes with its evidence, its risk and the effort it takes, ranked by dollars a month. Savings stay labelled projected until your bill confirms them.
Cloud and AI choices made without the numbers.
Compare equivalent services and machine types across AWS, Azure and Google Cloud, and price an AI provider’s usage export in your browser.
How it works
From a read-only connection to a ranked plan.
Connect, read-only.
Deploy our CloudFormation role — 82 read-only actions, listed one by one — or upload a FOCUS or CUR billing file. We publish the exact policy, nothing is installed, and you can delete the role any time.
See the bill explained.
Spend by service, account, team and region, what changed against the last period, and forecasts and budgets with their basis shown.
Act on what is worth it.
Findings ranked by monthly dollars, each with its evidence and its risk, and an owner where your tags name one. Your team decides what to change.
Four questions every cloud bill raises.
See how CloudJaeger answers each one, on a fictional estate where every figure adds up to the cent. Pick a question.Swipe the questions.
Sample estate · a fictional company on AWS billing, Sep 1 – 14 · every figure adds up to the cent
The thread: +$4,420 in 14 days → 84% of it is checkout‑api → a +$620/day step since Sep 9 → $18,600/mo at stake, projected
Overview
Sep 1 – 14, 2026 · compared with Aug 18 – 31 · billed cost · USD
Spend rose $4,420 (8.1%) against the prior 14 days. Three changes explain all of it — one service, checkout‑api, is 84% of the rise.
- Window spend
- $59,020Sep 1 – 14 · billed
- Change vs prior 14 days
- +$4,420+8.1% · checkout‑api explains 84%
- Forecast month-end
- ~$132,700projected · at the current rate
- Savings on a bill
- Nonenot measured on a bill yetHow verification works
Daily billed cost
Sep 1 – 30, 2026 · USD · axis starts at $0
- Actual · billed
- Projected (hatched)
Basis: billed cost, all charges, USD · Sep 15 – 30 projected at the current rate, $4,605/day (usual $3,900 + $705 of changes) · every point is in the data table.
What moved
share of the +$4,420
- checkout‑api · Fargate compute ×1.7+$3,72084%
- analytics‑warehouse · cross-AZ transfer+$1,06024%
- log‑archive · S3 storage down−$360−8%
- 18 other services · net $0 (+$56 / −$56)$00%
Rows add to the headline: +$3,720 + $1,060 − $360 + $0 = +$4,420.
Where it landed · by team
share of $59,020
- Payments$16,60028%
- Data Platform$14,61225%
- No owner tag$11,52220%
- Platform$9,59416%
- Catalog$6,69211%
Owner from the service tag on each resource; untagged spend stays visible, never spread.
- Billed through
- Sep 14 · 4 h ago
- Coverage
- 1 AWS payer · 21 services
- Basis
- Billed cost · all charges · USD
- Savings
- Projected only · none bill-verified yet
- Three causes that add up to the cent: +$3,720 + $1,060 − $360 = +$4,420.
- Month-end about $132,700 if nothing changes — hatched, because it is a projection.
- Every figure names its source and freshness: billed through Sep 14, synced 4 hours ago.
Overview · Cost Explorer · Architecture · Review
The walkthrough estate. The live demo is a different fictional estate — three clouds and 90 days of history — with its own numbers.
Open the live demo — no signupWhat CloudJaeger does
Cost, architecture and growth,
in one view.
Every feature listed under the three capabilities is marked Live, Preview, Demo-only or Planned, so you can see what works today.
Understand and reduce spend
Find the costs worth investigating.
Bring spend, ownership and opportunities into the same conversation. Give finance a clearer explanation and engineering a useful next step.
Compare periods and inspect spend by service, account, region or team. Review cost movements with the period and billing basis in view.
Connect spend to teams, products and cost centers. Keep configured, inferred and unknown assignments visible for showback and the next cost conversation.
Review idle-resource and capacity findings with their evidence, risk and effort. Give each next step an owner; projected savings remain estimates.
Explore coverage, utilization and commitment scenarios. Compare terms, scope, obligations and uncovered usage before considering a purchase.
A question to start with
Explore the cost example“Our bill increased. What should we investigate first?”
Compare the period. Inspect the service. Assign the review.Choose the right architecture
Put the design behind the spend in view.
Compare the same workload across clouds and size it to the evidence.
Compare supported cloud services and machine types against workload requirements. Review compatibility and operational trade-offs, priced at each cloud's baseline region, instead of assuming a direct replacement.
Review each rightsizing finding beside the utilisation it rests on and the rollback it would need.
Design findings and network cost for an AWS estate, each with its evidence and missing inputs.
Resources by account, region and network at list price, with what it cannot see listed. Swept nightly: Azure subscriptions and Google Cloud projects in full, AWS accounts in their billed regions, up to 20.
A reference design for a new workload, priced on all three clouds from dated list prices.
A question to start with
Explore architecture context“Can we improve this design without compromising the workload?”
Inspect the estate. Compare the options. Validate the trade-offs.Plan for growth
Understand the economics of what grows next.
Follow AI spend, budgets and forecasts with the measurement basis in view.
Follow instrumented usage by model, feature or tenant. Review token and GPU cost context, pricing assumptions and missing inputs.
Drop a usage export from your AI provider and see what it costs at list price. The file is read in your browser and never uploaded.
Compare actuals with budgets, inspect forecast assumptions and share scoped reports. Keep the reporting period, source coverage and export basis visible.
Cluster and workload usage with its cost basis: the bill for EKS with the CUR import, otherwise list price.
A cost per unit only where a matching denominator exists, never an invented one.
A question to start with
Explore AI cost management“What happens to our economics as usage grows?”
Choose a workload. Check the inputs. Review the growth scenario.Supporting your team
Reports & exports Team workflows Access controls Evidence & activity
View capability detailsBeta. Live works today; Preview works in part; Demo-only is shown on the sample estate; Planned is not supported yet. The grey note says where each was proven, or what it still lacks (status as of 2026-10-15). Savings and architecture scenarios are estimates until your bill confirms them.
Our approach
Confidence comes
from knowing the basis.
Useful cloud intelligence makes its evidence visible—and leaves consequential decisions with your team.
See the evidence policySources before certainty.
Billing, inventory and inferred relationships have different meanings. Their scope and freshness should travel with the finding.
Projected is not verified.
A modeled opportunity is an estimate. A saving needs billing evidence before it can be described as measured.
Your team stays in control.
Review evidence, operating risk and permissions before acting. Start with the published setup and security requirements.
Proof you can check
Check us, instead of trusting us.
We are in beta, so instead of customer logos, here is what you can check yourself. The figures are computed from the product’s own files, and each claim comes with the way to check it.
- 700
cloud services mapped across AWS, Azure and Google Cloud
122 with list prices · the rest listed without one - 3,780
machine types compared across the three clouds
virtual and GPU machines, with a graded equivalence verdict · Google Cloud: 12 of 533 priced, the rest listed - 82
read-only actions in the AWS IAM policy we publish
parsed from the CloudFormation template the connect wizard hands you - 20 Sep 2026 – 13 Oct 2026
when the list prices were read
public on-demand prices at each cloud's baseline region (us-east-1, eastus, us-central1)
No customers, certifications or savings figures are implied. The counts come from the product’s own modules — the service and machine catalogues, the read-only role’s CloudFormation template and the price book — and change only when those files do.
The AWS role we ask you to deploy is read-only, and listed action by action.
The list on /connect-security is parsed from the policy file as the page renders — 82 actions today. Count them yourself.
Open the permissions listThe file you drop into the AI cost audit never leaves your browser.
Open the network panel in your browser’s dev tools, then run an audit: no request carries the file.
Open the AI cost auditWe publish what each part of the product cannot do yet, beside what it can.
Every feature listed under the three capabilities carries its status — Live, Preview, Demo-only or Planned. /evidence adds where each one was proven and what it cannot do yet.
Open the evidence page
Your next cloud decision starts here.
Bring the whole
picture into view.
Applications are reviewed personally. Submitting one does not create an account.