Cloud cost management tools are best told apart by the job they do: reporting across clouds, allocation, budgeting and forecasting, anomaly detection, commitments, rightsizing, Kubernetes, cost estimation in code, policy, data pipelines and AI usage. Each type needs particular data and particular permissions, and for most of them the providers' consoles already ship a version for their own cloud. Knowing the type tells you which data has to exist first, which permissions a tool will hold, and where a second tool does more than AWS Cost Explorer, Azure Cost Management or Google Cloud's billing reports.
How it works
Every cost tool reads something: billing data, utilisation metrics or code. The types differ in what they read and whether they act. The right-hand column maps each type to the capabilities of the FinOps Foundation's FinOps Framework (CC BY 4.0), covered in the four FinOps domains and their capabilities.
| Type | Job | Data and permissions | In the provider consoles | Capability |
|---|---|---|---|---|
| Multi-cloud reporting | one view of every provider in one currency, with SaaS and AI invoices | each provider's billing export, read-only | per provider only | Data Ingestion; Reporting & Analytics |
| Allocation | cost per team, product or customer; shared cost split | export with tags, labels and account structure, plus your own rules | AWS Cost Categories, cost allocation rules in Azure, projects and labels in Google Cloud | Allocation; Invoicing & Chargeback |
| Budgeting and forecasting | plan, budget, variance | cost history, plan figures, business drivers; write access only to create provider budgets | AWS Budgets, budgets in Azure and Cloud Billing | Budgeting; Forecasting |
| Anomaly detection | flag unexpected spend | a daily cost series, read-only | AWS Cost Anomaly Detection, anomaly alerts in Azure, anomalies in Cloud Billing | Anomaly Management |
| Commitment management | size and buy reservations, Savings Plans and CUDs | hourly usage, existing commitments; purchase rights | recommendations for Savings Plans, reservations and CUDs | Rate Optimization |
| Rightsizing and scheduling | resize, stop and schedule resources | utilisation metrics, resource APIs; write access to act | AWS Compute Optimizer, Azure Advisor, recommenders in Google Cloud | Usage Optimization |
| Kubernetes cost allocation | cost per namespace, workload and label | cluster metrics for requests and usage, node prices; an agent in the cluster | split cost allocation data at AWS, AKS cost analysis, GKE cost allocation | Allocation |
| Estimation in code | the price of a change before deployment | Terraform plan and price lists; read access to code and pipeline | price calculators that do not read the code | Planning & Estimating |
| Policy engines | enforce tags, regions and sizes; clean up | resource APIs and inventory; write access for remediation | AWS tag policies, Azure Policy, Organization Policy in Google Cloud | Governance, Policy & Risk |
| Pipelines and BI | your own models and dashboards over the exports | exports in a warehouse; storage and query cost | the exports themselves | Data Ingestion; Reporting & Analytics |
| AI usage tracking | tokens and cost per model, key and feature | model providers' usage and cost APIs, request logs; admin keys | cloud-hosted models on the cloud bill | Allocation; Unit Economics |
For Kubernetes allocation, estimation in code and policy, open-source cost tools covers one open-source project each; Microsoft's AKS cost analysis is itself built on one of them, OpenCost. Pipelines and BI are the subject of cost data in spreadsheets and BI, budgets and rules as code of cost as code, assistants over cost data of AI assistants and MCP. How the model providers report usage and cost through their own admin APIs is covered in AI provider billing.
The Framework treats tooling itself as a capability, Automation, Tools, & Services, in the domain Manage the FinOps Practice. There it sorts tools by breadth, from one that spans nearly every capability to one built for a single capability, and treats automation and professional services as routes of their own. Its approach starts with an inventory: what is already in use is laid beside the capabilities, and whatever no tool covers is the gap. Among the selection criteria the Foundation counts interoperability and open standards. Its FinOps Landscape filters entries by type (tool, services, training), technology category, data provider, capability, specialisation, certification and compliance, and by whether a vendor has adopted FOCUS. Its technology categories go beyond the Framework's five described in FinOps scopes and technology categories: on 2026-10-16 it also filters by Licences and Private Cloud.
What it means for your cost
A provider's console sees only that provider. Microsoft describes Cost Management as a tool for Microsoft Cloud costs, and a Google Cloud Billing account covers Google Cloud, Google Maps Platform, Firebase and Google AI Studio. A second tool starts where a question crosses that line: several clouds, SaaS and AI invoices, revenue or user numbers for unit economics, or one reporting currency, as multi-cloud reporting in one currency shows. How the term relates to FinOps is set out in cloud cost management and FinOps compared.
Permissions follow the type, in three steps. Reading billing data is enough for reporting, allocation, budgets and anomalies, and each provider has a role for it: Billing Account Viewer at Google, Cost Management Reader at Azure, and at AWS IAM policies that can be limited to read actions such as ce:GetCostAndUsage. Reading resources and metrics is what rightsizing, Kubernetes allocation and policy need. Write access belongs only to tools that act: resize, stop, buy commitments, remediate. Which role covers which job is compared in cost data access control.
Every tool raises the same security questions:
- Which role on which scope, and is it limited to reading?
- How is access granted, and how is it withdrawn when the tool goes?
- Where is the data stored and processed (data residency)?
- Who owns each write action, and where is it logged?
What a small team takes from the 2026 Framework before choosing any tool is covered in FinOps Framework 2026: what small teams should actually change.
Where it goes wrong
- The type before the data. An allocation tool is bought before tags and account structure exist, and mostly shows unallocated cost. Anomaly detection needs a daily cost series, Kubernetes allocation needs resource requests that are set.
- Two tools, two cost measures. The console and a second tool show different totals for the same month, because one is amortised and the other billed, or they differ in data cut-off or currency. Both are called "the cost".
- Savings counted twice. The console and a second tool recommend the same commitment, and both estimates end up in the plan. The consoles also differ: Azure accounts for existing reservations and savings plans, AWS for existing Savings Plans, while Google's FinOps hub ignores CUDs already bought when estimating a single recommendation.
- Write access nobody owns. A schedule stops a database overnight that another team needs, and nobody knows who set it up. The permissions from the pilot remain after the tool has been replaced.
- Read access wider than the job. A reporting tool is given an administrator role because that was quicker to grant than a read-only one.
Sources
Read on 2026-10-16. FinOps Foundation FinOps Domains, Automation, Tools, & Services, Technology Categories and FinOps Landscape; AWS Cost Explorer, AWS Budgets, Cost Anomaly Detection, Cost Categories, Savings Plans recommendations, Compute Optimizer, split cost allocation data, AWS Pricing Calculator, tag policies and IAM policies for Cost Management; Microsoft Cost Management overview, assigning access to cost data, Advisor: resizing or shutting down underused VMs, reservation recommendations, AKS cost analysis, Azure Policy and pricing calculator; Google Cloud Billing overview, Cloud Billing access control, budgets, cost anomalies, Billing Reports, CUD recommendations, FinOps hub, Active Assist, GKE cost allocation, Organization Policy and pricing calculator; OpenAI Admin API costs endpoint; Anthropic Usage and Cost API.