Kubernetes Costs page

Understand cluster, namespace, workload and pod costs, idle capacity, rightsizing, carbon and health.

Kubernetes Costs page

See what each cluster, namespace, workload and pod costs, how much of it is idle, and where requests are bigger than real usage.

Where in CostGlassCosts → Kubernetes Costs

PlanGrowth and above (visible but greyed out on lower plans)

Who can do thisEveryone in the organisation can view it

Time needed20 minutes to read, then use it as a reference

In short

Kubernetes costs are an allocation of your cloud bill: they show who inside the cluster uses what, and are not added to your total cloud spend.
The figures are a monthly run-rate projected from the latest push (about 15 minutes of measurement), not a month of measured spend. They move as your workloads change.
Idle cost is the price of node capacity that no pod has requested. Efficiency compares what pods use with what they asked for.
Eleven tabs under "Allocate By" change how the same data is grouped: Namespace, Workload, Pod / Container, Cluster, Team / Label, Cluster Assets, Cost By Type, Rightsizing, Network Costs, Carbon and Health.
Use it to find over-sized requests, empty node capacity and the namespaces that should be charged back to teams.

Before you begin

At least one cluster connected and pushing data: Connect a Kubernetes cluster.
A plan that includes Kubernetes Costs (Growth or above).
For the Health tab, the optional read-only RBAC from the install manifest. For the Carbon tab, OpenCost installed with the carbon option.

What OpenCost measures

OpenCost watches the cluster and prices what each workload uses. The page can show these cost types; each is a field OpenCost itself reports and none of them is estimated by CostGlass:

Cost type	What it is
CPU	The cost of CPU the workload reserved (requested) or used, whichever is greater, at your cloud's price.
RAM	The same for memory.
Storage	Persistent volumes (disks) the workload claims.
Network	Data transfer. Cross-zone, cross-region and internet egress are split out on the Network Costs tab.
Load Balancer	Load balancers created by Kubernetes services.
GPU	GPU time. The GPU swatch is hidden unless your data has GPU cost.
Shared	Cost of shared services that OpenCost spreads across workloads. The exporter treats the kube-system, istio-system and monitoring namespaces as shared.
External	Costs outside the cluster that OpenCost attributes to workloads (this is empty unless your OpenCost setup reports it).
Idle	Node capacity that no workload has claimed.

OpenCost also reports, for each workload, how much CPU and memory it requested and how much it actually used on average. That pair is what makes rightsizing possible.

Namespace: A named group inside a cluster, often one per team, product or environment.
Workload: A running application managed by Kubernetes: a Deployment, StatefulSet, DaemonSet or Job.
Pod: The smallest running unit: one or more containers that run together on a node. A workload runs one or more pods.
Node: A machine (virtual or physical) the pods run on. This is what your cloud bills you for.
Request: The amount of CPU or memory a pod asks Kubernetes to reserve for it. Reserved capacity is paid for whether or not it is used.
Idle / slack: Capacity on the nodes that no pod has requested.

Why numbers are a monthly run-rate, and what that means

OpenCost reports each cost for the minutes it measured, normally one 15-minute step. The exporter sends a fresh measurement every 15 minutes and each push replaces the previous one. If CostGlass showed that raw figure it would be around 2,900 times too small, so CostGlass projects it: it merges the workload across the measured windows and scales the cost to a month (730 hours).

Fact	What it means for you
Figures are a projection from the latest measurement	If something is unusually busy or quiet in the last 15 minutes, the whole month projection moves with it. Look at trends over several days before deciding.
A push under 5 minutes of data is not projected	Too little data to project from, so the figure is stored as sent.
Only the latest push is kept for live allocations	The tables on the page show the latest snapshot. Each push replaces the previous one, so there is no per-pod history inside CostGlass.
Daily history is kept at namespace level	Each push also writes a daily cost row per namespace, using the run-rate day, plus one for idle capacity when node prices are known. These feed the trend chart.
The time-range buttons rescale the current run-rate	The summary and table figures use the latest monthly run-rate multiplied by the length of the range divided by 30 days (for example 7D is 7/30 of a month, 90D is three months). They are not sums of actual past spend.

Treat this as a guide to where cost sits, not an invoice

The page tells you how a cluster's cost is shared between teams and where waste sits. For the amount you were actually billed, use Cost Explorer on your cloud account.

Node pricing, idle cost and reconciliation to your bill

OpenCost's pod figures only cover what pods requested. To show what the machines cost (and so what is idle), CostGlass prices each worker node when it can.

Cloud	How nodes are priced
AWS (EKS)	From CostGlass's approximate list-price table for the instance type.
Azure (AKS)	From CostGlass's approximate pay-as-you-go price table for the VM size.
Linode (LKE)	From Linode's own published plan prices.
Google Cloud, Oracle, DigitalOcean and others	No node price source in CostGlass, so nodes are not priced.

All or nothing. The cluster is priced by its nodes only when every node can be priced. If even one node cannot, the cluster stays "pod-priced": the total is just what OpenCost allocated to pods, and the idle figure is not rebuilt from node prices. CostGlass does not guess a node price.

When all nodes are priced: cluster cost is the sum of node prices; each node's idle cost is its price minus the pods scheduled on it (never below zero); and the cluster efficiency is cost not idle divided by cost.

Reconciliation: the coloured pill under the intro text

Where your cloud billing data is connected, CostGlass compares the Kubernetes node figures with what the cloud actually charged for those nodes. Each cluster gets one pill:

Pill	Meaning
Reconciled to your bill	Every node was found in your billing data. Pod shares are scaled so they add up to the billed node cost.
Partly reconciled	Some nodes were matched. Hover the pill to see how many matched and why the rest did not.
Estimated at list price	No matching bill data yet, or the cloud does not support matching. The numbers are list-price estimates and the pill says why.

AWS matching uses EC2 instance IDs in your billing data (CUR with resource IDs). Azure, Google Cloud, Oracle and Alibaba matching also exists, but was written from documentation and sample data and has not been run against a real export from those clouds, so expect "Estimated at list price" for them until proven. Linode and DigitalOcean always stay at list price. Billing data is normally about a day behind. Reconciliation runs automatically after each push; owners and admins can also trigger it through the API.

Page layout and controls

Kubernetes Costs: KPI cards, the Allocate By tabs and controls, the spend trajectory chart and the cost breakdown table.

From top to bottom: the header (with Ask CostGlass AI, the connection status and Refresh Telemetry), a one-line reminder that Kubernetes costs are an allocation, a reconciliation pill per cluster, four KPI cards, the controls card, the spend trajectory chart, rightsizing and slack cards, and the main table for the selected tab.

Control reference

Control	What it does	Options / default
Allocate By	Chooses how the chart and table are grouped and which view is shown.	Namespace (default), Workload, Pod / Container, Cluster, Team / Label, Cluster Assets, Cost By Type, Rightsizing, Network Costs, Carbon, Health
Cluster	Limits everything to one cluster.	All Clusters (default) or any connected cluster
7D / 30D / MTD / 90D / 6M / Custom	Sets the time range used to scale the figures and to draw the chart.	MTD (month to date) by default. Custom shows From and To date boxes and an Apply Range button, with Reset to go back.
USD / EUR / GBP	Currency used to show money.	Your organisation's setting
Copy link	Copies a link that reproduces this exact view (tab, range, cluster, label key, filters) for anyone in your organisation.	-
CSV	Downloads the current table as a CSV file.	-
Refresh Telemetry	Re-reads fresh data from CostGlass, ignoring the page's saved copy.	-
Group by label key	Only on the Team / Label tab: which Kubernetes label to group cost by.	team (default), app, env, service, component
Filter ... box	Free-text filter on the table rows.	Placeholder is "Filter workloads..." or "Filter namespaces..." and so on
All kinds	Only on the Workload tab: shows only Deployments, StatefulSets, DaemonSets, Jobs and so on.	All kinds (default)
Rows per page	Table page size.	10, 15, 25, 50, 100
Chart type switcher	Changes the trajectory chart.	Stacked bars (default), Lines, Share (treemap)
Connect OpenCost Agent / Config	Opens the connect window with the install commands. After a cluster is connected it is shown as Config.	-

Saved copy and refresh

To load quickly, the page keeps the last result of each view in your browser for 15 minutes. The page also asks the browser-wide cache for results for up to an hour. Refresh Telemetry skips both and fetches fresh data. The list of connected clusters is always read fresh. New data from the cluster arrives at most every 15 minutes anyway, so refreshing more often than that will show the same numbers.

The four KPI cards

Card	How it is worked out	How to read it
Total Kubernetes Spend	Sum of each cluster's monthly run-rate, scaled to the selected range. The line below shows the number of clusters and nodes (labelled EC2 Nodes, AKS Nodes, GKE Nodes and so on by provider).	Your container cost under measurement. It is an allocation and is not added to your total cloud spend.
Allocated Spend (Requests)	Total spend minus idle spend. The line below shows the percentage "guaranteed to pods".	Capacity that pods have reserved. High is good: capacity you pay for is claimed.
Idle & Slack Capacity	Idle cost summed from the namespace rows, with a percentage of the total.	Capacity nobody has claimed. This is the first place to look for savings.
Cluster Efficiency Score	The average of each cluster's efficiency, shown in green from 80%, amber from 60%, red below. The line below shows "+£x/mo savings found", which is 65% of idle cost.	Higher is better. The savings figure is a planning number: some headroom should always stay. It does not appear for clusters that have no efficiency data (a missing score is treated as 75).

Note

The AI Advisor link on the efficiency card opens CostGlass AI, which can explain these figures using your page's data.

The spend trajectory chart

The chart title follows your tab, for example Namespace Spend Trajectory. It shows daily bars (or monthly for long ranges) split by the grouping you selected, drawn from the daily history CostGlass has recorded. Click a legend item to focus on it, and Clear selection to go back. Switch to Lines to compare trends, or Share (treemap) to see proportions at a glance.

Tip

If the chart is empty, the cluster has probably not been connected for more than a day yet. The text under the chart says so. The table below is still live.

Request vs Usage and Slack

The right-hand card shows two bars, Compute CPU Allocation and Memory RAM Allocation, each split into three parts:

Colour	Part	Meaning
Green	Utilized	Capacity pods are really using.
Blue	Headroom	Capacity pods reserved but are not using. The gap between request and usage.
Amber	Slack	Capacity on the nodes that no pod has reserved. This is the idle share.

These bars are a visual summary

The Slack share is real (it comes from the idle share of spend). The Utilized part is scaled from the efficiency score for display, so use the Rightsizing tab for real request and usage numbers per workload.

Below the bars, ... / month in Cluster Slack shows the idle cost, with advice that matches your cloud (for example Karpenter consolidation or Spot instances on AWS, Cluster Autoscaler on AKS, GKE and LKE).

The left card, Workload Rightsizing Recommendations, lists workloads whose efficiency is under 60% and whose idle cost is at least $2 a month. Each shows the workload, namespace, cluster, the estimated monthly saving (equal to its idle cost, an upper bound), the reason, and the current CPU request. Its Quick Apply tag is a label only: CostGlass never changes your cluster.

The Allocate By tabs

Namespace, Workload, Pod / Container, Cluster, Team / Label

These five tabs show the main table, Cost Breakdown by ..., with these columns:

Column	Meaning
#	Rank by cost.
Name	The namespace, workload, pod, cluster or label value.
Context / Cluster	Where it lives: its cluster and, for workloads and pods, its namespace.
Kind / Replicas	Workload tab only: Deployment, StatefulSet, DaemonSet, Job and how many replicas.
Node Target	Pod tab only: the node the pod runs on.
Share of Spend	Its percentage of the total for this view, with a bar.
Efficiency	Used versus requested. Shown as a badge: Optimal from 80%, Fair from 60%, Waste Alert below 60%.
Cost	Its monthly run-rate (for the selected range).

Team / Label groups by a Kubernetes label you choose. The label keys offered are team, app, env, service and component. Cost under a label value is only as good as your labelling: anything without the label is grouped separately. The exporter sends the app, env and team labels, so those three are the ones most likely to have data.

Cluster Assets

Shows Infrastructure Assets Breakdown: summary tiles (Total Asset Spend, Compute Nodes, Provider & Region, Cluster Efficiency) and a table withComponent Kind, Name / Identifier, Cluster, Specification / Details, State, Cost Basis and Monthly Cost. This is the view where node prices and the control plane show up. The Cost Basis column says how each price was obtained, for example a list price table.

Cost By Type

Splits cost into CPU, RAM, Storage, Network, Load Balancer, GPU, Shared, External and Idle, grouped by namespace, workload or cluster (switch at the top of the card). A zero is a real zero, not a missing value. Use it to see whether cost is driven by compute, storage or something unexpected like load balancers.

Rightsizing

Request vs Usage (Rightsizing) shows, per workload, CPU Request / Usage and Memory Request / Usage next to Monthly Spend, with a filter box. Usage is OpenCost's average over its measured window, so it appears only once Prometheus has collected enough history. If a workload shows nothing, it has not reported usage yet; the page does not invent a figure. AI Rightsizing Insights asks CostGlass AI to review the table.

Network Costs

Splits network cost into Cross-Zone, Cross-Region, Internet Egress and Other. OpenCost reports these only when network cost allocation is turned on in OpenCost inside your cluster. That is a setting in your OpenCost deployment that CostGlass cannot change for you, so the tab may be empty until you enable it. Cross-zone and egress traffic is often a hidden cost: services that talk across availability zones can be moved closer together.

Carbon

Carbon Emissions shows kilograms of CO2e. It comes from OpenCost's own carbon estimates (which use Cloud Carbon Footprint coefficients for your cloud, region and instance type), not a CostGlass estimate. OpenCost reports carbon per machine and disk, so CostGlass sums it for each cluster. There is no namespace or pod split.

The exporter measures a one-hour window. Projected Monthly CO2e is that hourly figure multiplied up (daily is hourly times 24, monthly is times 720). It is a projection, not a measured total.
A cluster that has never reported carbon shows not reported, never 0. That means OpenCost has no carbon data for it, usually because the carbon option was not turned on.
The small bar chart Daily CO2e Trend (projected) shows the history of those projections.

Health

Cluster Health uses a direct read of the Kubernetes API (this is why the optional RBAC is needed), because cost data has nothing about pod health. Each cluster gets a score out of 100 using a published formula:

Item	Effect on the score
Failed pods (including pods stuck in CrashLoopBackOff or ImagePullBackOff)	Minus 50 times the share of pods that are failed
Pending pods	Minus 15 times the share of pods that are pending
Nodes not Ready	Minus 25 times the share of nodes not ready
Recent restarts	Minus the number of recent container restarts, up to 10
Pending storage claims	Minus 2 for each, up to 10

The overall score is the average of the clusters that reported. Colours: green from 90, amber from 70, red below. Each cluster line shows pods healthy out of total, failed, pending, nodes ready, recent restarts and pending storage claims. Kubernetes only keeps a running total of restarts, so recent restarts is the increase since the previous push (never negative; a recreated pod resets its own counter).

Making decisions from the page

What you see	What it usually means	What to do
High Idle %, with low efficiency	Nodes are bigger or more numerous than the pods need.	Reduce node count or node size, use a cluster autoscaler or consolidation tool, and consider Spot or preemptible capacity for tolerant workloads.
A workload with large Headroom (request far above usage)	Requests were set generously and never tuned.	On the Rightsizing tab, lower the CPU and memory requests step by step toward actual usage plus a safety margin, and watch for throttling or out-of-memory restarts.
Efficiency under 60% on a workload that costs little	Not worth your time.	Sort by cost and tackle the biggest idle cost first.
A workload at 10% efficiency or lower	Possibly abandoned or no longer needed.	Check with its owner and remove it if it is not needed. CostGlass flags these in Savings.
Many half-empty nodes	The scheduler cannot pack pods tightly (spread rules, large requests).	Look at bin-packing: smaller requests, fewer spread constraints, or different node sizes.
Spend that jumps between days	Run-rate sensitivity, a scaling event or a deployment.	Compare days on the chart before concluding; use Cost Explorer and the anomaly page for real billed changes.
Cross-zone or egress cost	Chatty services spread over zones or regions.	Place related services in the same zone where reliability allows.
Health score dropping, many restarts	Pods crashing, nodes not ready or storage not binding.	Investigate with kubectl before cutting resources: do not rightsize unhealthy workloads.

Tip

Do not lower requests only by looking at one 15-minute snapshot. Check usage over a full business cycle (at least a week) before reducing a request.

Budgets, Savings and Reports

Budgets. You can create a budget for Kubernetes (OpenCost), scoped to all clusters or one cluster, and optionally one namespace. OpenCost overwrites its figures on every push, so there is no month-to-date total. The budget is therefore judged on the projected month, and the card says "tracked on projected run-rate". See Budgets.
Savings. The savings page can list Kubernetes recommendations built only from OpenCost's own figures: workloads at or below 10% efficiency (saving is the whole workload cost), over-requesting workloads (saving is the reported idle cost, an upper bound), and cluster slack (an estimate that assumes half of the slack can be reclaimed, stated on the card). Kubernetes savings are grouped with the host cloud bill so the same money is not counted twice. See Savings.
Reports.Reports can group by Kubernetes Namespace (OpenCost) and has an Include Kubernetes OpenCost Allocations switch. This reads the latest snapshot only, so it has no previous period to compare. Saved filters do not apply to Kubernetes namespace allocations.

Worked recipes

Charge back cluster cost to teams by namespace

Open Kubernetes Costs and choose Namespace under Allocate By. Set the range to MTD.
Check the reconciliation pill. If it says Reconciled to your bill, shares are scaled to what the cloud billed for the nodes. Otherwise they are list-price estimates.
Note that Idle is not owned by any team. Decide a policy: spread it in proportion to each team's share, or keep it as a platform cost.
Click CSV and send the table to finance. For a repeatable report, use Reports with the Kubernetes Namespace grouping.

Remember the figures are a run-rate. For month-end numbers that must tie to an invoice, base them on your cloud bill.

Find the most wasteful workload

Choose the Workload tab and look at the Efficiency column. Click through to see the lowest values.
Open the Rightsizing tab for the CPU and memory request against usage of those workloads.
For each, compare the gap in cost terms using the Monthly Spend column, and start with the largest.
Lower requests towards usage plus a margin, redeploy, then check again after a week.

Is the cluster too big?

Choose Cluster and read Idle & Slack Capacity and the efficiency score.
Open Cluster Assets to see the nodes and what each costs.
If idle is large and stable, remove or resize nodes. If it follows a daily pattern, use autoscaling.

Honest limits

Figures are a monthly run-rate projected from about 15 minutes of measurement, not measured monthly spend.
Live allocations are one snapshot. There is no per-pod history inside CostGlass; only the daily namespace-level and idle history exists.
Node prices exist only for AWS, Azure and Linode. For Google Cloud, Oracle and others, nodes are not priced and clusters are pod-priced (no node-based idle).
AWS and Azure node prices come from approximate in-product tables, not your actual rates (reservations, Savings Plans and discounts are not applied unless the cluster is reconciled).
Reconciliation to the bill is tested for AWS; for Azure, Google Cloud, Oracle and Alibaba it has not been confirmed against real exports, and Linode and DigitalOcean are not reconciled.
Carbon is a projection from a one-hour window. It is reported per cluster only.
Network costs depend on a setting in your OpenCost deployment. Usage figures need Prometheus history.
Health needs the optional read-only RBAC.
CostGlass reads and reports. It never changes your cluster.

Frequently asked questions

Why is my Kubernetes total different from my cloud bill?

Kubernetes figures are an allocation of the bill using list prices and a short measurement window. They are not added to your total spend, and they only match your invoice when the cluster is reconciled.

Why does MTD show the same number as 30D?

Both scale the current monthly run-rate: MTD uses one month and 30D uses 30 days divided by 30, so they are equal. The charts use the daily history and differ.

Why is a cluster's idle cost zero or missing?

Idle capacity is built from node prices. If any node cannot be priced (for example on Google Cloud or Oracle), the cluster stays pod-priced and no node-based idle is calculated.

Why does Rightsizing show no usage?

OpenCost reports usage only once Prometheus has scraped enough history. Wait a while after installing, and check Prometheus is healthy.

Why is Carbon "not reported"?

OpenCost has not sent carbon data for that cluster. Install or upgrade OpenCost with the carbon option and re-apply the manifest.

Does Quick Apply change my cluster?

No. It is a label. Changing requests is done by your team in your own manifests or pipelines.

Can I see a single pod's cost history?

No. Only the latest snapshot is kept for pods. Daily history exists at namespace level.

Who can see this page?

Everyone in your organisation on a Growth or higher plan. Kubernetes figures are organisation-wide: they are not narrowed by organisation-hierarchy scopes.

All Documentation Guides

Welcome to CostGlass — What CostGlass does, who it is for, how your data gets in, the plans and regions, and a plain-English glossary.
Quickstart: your first hour — Sign up, connect a first cloud, invite your team, set a budget and turn on anomaly alerts, with a first-week checklist.
Finding your way around — A tour of the sidebar, top bar and page controls: what every page answers, plan gating, currency, language, links and caching.
Dashboards — The Overview dashboard widget by widget, plus building custom dashboards you can share with your organisation.
Cloud Resources — The multi-cloud inventory: filters, search, categories, estimated cost, terminated resources and CSV or Excel export.
Uptime, incidents & support requests — Monitor public endpoints, route alerts, see incidents and raise requests to CostGlass engineers.
How CostGlass accesses your data — Security and data ingestion architecture: how CostGlass reads multi-cloud billing and telemetry through read-only roles and least-privilege permissions.
Dashboard & Platform Screenshot Tour — Visual walkthrough of the CostGlass interface with screenshots covering multi-cloud spend, Cost Explorer, anomalies, Kubernetes allocation, and savings.
Connecting your clouds: overview — Every cloud and service you can connect, how each one connects, what access it needs, how long data takes and your plan limits.
Connect AWS (CloudFormation) — Deploy the generated CloudFormation template, choose organization or single account and optional features, then verify.
Connect AWS with Terraform instead of CloudFormation — A Terraform equivalent of the generated AWS template, with an Organization variant and a way to compare it to the wizard.
Connect Azure — Admin consent for the CostGlass application, Reader and Cost Management Reader roles, management group or subscription scope, and a Terraform version.
Connect Google Cloud — Enable the FOCUS BigQuery export, run the IAM grants CostGlass prints, verify, and a Terraform version.
Managing connected accounts — The Cloud Accounts page: statuses, syncing, errors, naming, disconnecting, and the FOCUS 1.0 and CUR Export pages.
Connect Oracle Cloud (OCI) — Give CostGlass read-only access to Oracle Cost and Usage Reports with a Resource Manager stack or your own Terraform.
Connect Alibaba Cloud — Connect Alibaba Cloud through Bill Subscription files in OSS and a read-only RAM role created by a ROS template.
Connect DigitalOcean — Connect DigitalOcean accounts with a read-only personal access token and see invoices by month.
Connect Akamai / Linode — Connect Linode (Akamai Connected Cloud) accounts with an account read-only token.
Connect Cloudflare — Connect Cloudflare usage-based charges with a Billing Read account token and see credits and unpaid debt.
Connect OpenAI, Anthropic, Datadog and Snowflake — Connect AI and software vendors with admin keys or a Snowflake key pair so their spend sits beside your cloud costs.
Cost Explorer — Slice, compare and drill into spend across every connected cloud: group by, filters, charts, overlays, details, AI, links and export.
Saved views, saved filters & scheduled emails — Keep, share and email the Cost Explorer views you use often, and reuse include/exclude filters across views and reports.
Reports — Build, compare, export, save and schedule spend reports, with presets, Excel, CSV and PDF exports.
SaaS & AI Spend — Spend and token analytics for OpenAI, Anthropic, DeepSeek, Datadog, Snowflake and cloud AI services.
Unit Economics — Cost per business unit: define metrics, send volumes, set targets and alerts.
Connect a Kubernetes cluster — Install OpenCost and the CostGlass exporter in your cluster so cost data is pushed to CostGlass every 15 minutes.
Kubernetes Costs page — Understand cluster, namespace, workload and pod costs, idle capacity, rightsizing, carbon and health.
Tagging strategy: getting the most from tags — How to tag cloud resources, enforce tagging, and when to use cloud tags or virtual tags in CostGlass.
Tag Explorer — Group billed spend by the value of a tag, see tag coverage per cloud and drill into resources.
Tag Compliance — Define required AWS tags, measure compliance, and fix missing tags with Quick Tag, batch apply or a CLI script.
Virtual tagging — Allocate spend by team, product or environment with rules, no cloud changes, with a rule builder and CostGlass AI.
Forecasting — Forecast month-end spend, the next months and a yearly run-rate, and simulate the cost of changes.
Budgets — Create spending limits for a cloud, account, service or team, and see how CostGlass budgets differ from native AWS, Azure and Google budgets.
Anomaly detection — How CostGlass finds unusual daily cost, how to tune it, explain it and be notified.
Cost alerts — Alert when spend rises versus the previous day, week, month or quarter.
All alert and notification types — A comparison of every alert and notification in CostGlass, with channel setup and testing.
Issues & annotations — Dated notes on spend, and a lightweight issue tracker for cost work with Jira push.
Savings playbook — A prioritised, step-by-step programme for reducing your cloud bill, with a 30/60/90-day plan.
Savings recommendations — How the Savings page finds, prices and tracks savings for every connected cloud and platform.
Rightsizing — Find servers and databases that are too big, or too small, using real CPU and memory measurements.
Commitments: Savings Plans & Reserved Instances — Track coverage, utilisation, expiry and purchase recommendations for Savings Plans, Reservations and committed use discounts.
Instance Scheduler — Start and stop non-production servers on a weekly schedule to stop paying for idle hours.
Giving your team access — Invite people, choose Owner, Admin or Member roles, remove users and limit what they see.
Setting up your organisation hierarchy — Build a tree of nodes, place cloud accounts by hand, rule or CSV, and scope people to a branch.
Designing your hierarchy (with enterprise examples) — Principles and worked patterns for AWS Organizations, single accounts, multi-cloud groups, MSPs and regulated regions.
Single sign-on (SAML & OIDC) — Connect Okta, Entra ID, Google Workspace or another provider, verify your domain, test and require SSO.
API keys & AI assistants (MCP) — Create API keys and connect Claude Desktop, Cursor or Claude Code to your cost data.
Audit log — See who changed what and when, with before and after values, kept for 24 months.
Data residency & retention — Your fixed region, how long cost history is kept, purge now and the data handling summary.
Notification settings, Jira & chat — Weekly spend summary, Slack and Teams webhooks, and connecting Jira Cloud.
Troubleshooting & Common Setup Fixes — Solutions for common cloud connection and ingestion issues across AWS, Azure, Google Cloud, OCI, Alibaba Cloud, Kubernetes, and SSO.

Kubernetes Costs page