Skip to content
Case study · Cloud cost & security

CloudMonitor: one read-only view of AWS, Azure and Google Cloud

Teams running more than one cloud had three consoles, three bills and no single answer to "what do we run, what does it cost and what is exposed?". We built a read-only dashboard that answers all three without asking for write access to anything.

Laptop showing a cloud analytics dashboard
Case study at a glance
  • 3 cloudsAWS, Azure, Google Cloud
  • 17 checksGraded A to F
  • 13 AWS typesInventoried resources
  • 28-day baselineSpend anomaly scoring

The problem

Cost and security tooling is either cloud-specific or priced for large enterprises. Smaller teams on two or three clouds checked each console by hand, found idle resources months late and learned about an open storage bucket from someone else.

What we built

A multi-tenant Laravel application that connects to each cloud with read-only credentials, inventories resources, runs security checks, recommends cost fixes and flags unusual daily spend.

Where it stands

Live at cloud-monitor.radiatus.com with a free tier. We use it ourselves and in client cost reviews to see where a bill is going before changing anything.

Numbers you can check

Every figure on this page comes from the running product or its source code, checked on 7 November 2026. None is a client result.

FigureWhat it countsWhere it comes from
3Cloud providers behind one provider contract: AWS, Azure, Google CloudThe application's provider implementations
17Security checks: 9 on AWS (S3, EC2, security groups, RDS, IAM), 4 on Azure, 4 on Google CloudThe compliance engines per provider
13AWS resource types inventoried, from EC2, EBS and RDS to Lambda, DynamoDB, load balancers, Route 53 and VPCsOne reader per resource type
28 daysTrailing baseline each day's spend is scored against (z-score, default threshold 2.0)The anomaly detector
3Alert types sent by email: monthly budget, new critical finding, stale access keyThe alert rule model
A–FSecurity grade: 100 minus severity-weighted penaltiesThe scoring service

Read-only, on purpose

The first design decision was the one that makes a monitoring tool safe to connect: CloudMonitor never needs permission to change anything. On AWS it uses read-only roles and calls APIs such as Cost Explorer, CloudWatch metrics and CloudTrail lookups. On Azure it authenticates with client credentials against Azure Resource Manager and queries Cost Management. On Google Cloud it signs its own short-lived tokens. A leaked CloudMonitor credential can read an inventory; it cannot delete a database or open a port.

One contract, three clouds

Each cloud sits behind the same provider contract: list resources, fetch costs, run checks. AWS uses its official PHP SDK. Azure and Google Cloud are called over their REST APIs directly, which kept dependencies small and made the code that handles authentication easy to audit. Adding a fourth provider means implementing the contract, not touching the dashboard.

Scheduled jobs sync inventories, evaluate alert rules hourly and build daily reports. Every record is scoped to a tenant, so data from different organisations never mixes, and one organisation can connect several cloud accounts.

Spend anomalies without false alarms

Simple "spend went up 20%" rules fire every month-end and every launch. CloudMonitor scores each day's spend against a trailing 28-day baseline using a z-score, and ignores jumps smaller than one dollar so tiny accounts do not flap. Days already flagged are kept out of later baselines, so a single spike cannot raise the bar and hide the next one. The result is shown on the dashboard next to the services that drove it.

Security checks people act on

The 17 checks focus on the findings that cause real incidents: public storage, open security groups, unencrypted databases and disks, and stale or over-privileged IAM keys. Each finding has a severity, and the account grade is 100 minus severity-weighted penalties, from A to F. The checks are modelled on CIS-style controls; the product does not claim a formal CIS Benchmark mapping.

The stack

LayerChoiceWhy
ApplicationLaravel 13 on PHP 8.3A mature framework with queues, scheduling and policies built in
AWSAWS SDK for PHP: Cost Explorer, CloudWatch, CloudTrail and resource APIsOfficial SDK with read-only roles
Azure and Google CloudREST APIs with client-credentials and signed-JWT authenticationFew dependencies, auditable auth code
Access controlSanctum, role and permission package, activity logWho saw and changed what, per tenant
Front endVite and Tailwind CSSFast dashboards without a separate SPA

What it does not do yet

  • Google Cloud spend is not shown yet: it needs a BigQuery billing export, which the product does not read today. Google Cloud inventory and checks work.
  • Spend anomalies appear on the dashboard; they are not sent as email alerts. Email alerts cover budgets, new critical findings and stale keys.
  • Checks are CIS-style, not a certified CIS Benchmark assessment.

What this means for your cloud bill

Most cloud cost work starts with the same three questions CloudMonitor answers: what is running, what is it costing, and what is exposed. In a cloud infrastructure or cost engagement we answer them for your accounts first, then fix the waste and the risky findings as code. Ongoing operations, including monthly cost reviews, are part of our managed services.

Our FinOps playbook explains the method behind the cost recommendations.

FAQ

Questions about this case study

AWS, Microsoft Azure and Google Cloud. Inventory and security checks work on all three; spend reporting works on AWS and Azure, and Google Cloud spend needs a BigQuery billing export that is not supported yet.

No. It uses read-only credentials on every provider, so it can list resources and costs but cannot change or delete anything.

Each day's spend is scored against a trailing 28-day baseline using a z-score, small absolute jumps are ignored, and flagged days are kept out of later baselines. Anomalies are shown on the dashboard.

Seventeen: nine on AWS covering storage, compute, security groups, databases and IAM, four on Azure and four on Google Cloud. Findings roll up into an A to F grade.

Yes. We often start a cost or security review by connecting CloudMonitor read-only, then turn the findings into a prioritised fix list and make the changes as infrastructure code.

Work with the team that built it

Need something like this for your organisation?

Tell us what you are connecting, running or securing. The engineers behind this case study reply personally.

  • A senior engineer reads every enquiry
  • Happy to sign an NDA first
  • You get a written next step, not a sales sequence
Last updated