finops-architect-toolkit
FinOps calculators for AWS architects: RI, S3, Lambda and tagging — in one conversation.
git clone https://github.com/fernando-moretes/ref-finops-architect-toolkit.gitListen to guide
generated on playGenerated only on first play
Powered by Amazon Polly + OmniVoice
A set of FinOps calculators and templates built on Next.js 16 for architects who need to defend an AWS cost decision with a number on the table — RI versus On-Demand, S3 storage class, Lambda cost and tagging strategy.
Why this repository exists
After years reviewing architectures in financial environments, the cloud bill always reached my desk before it reached finance. The reason is simple: architecture choices — instance type, storage class, Lambda memory, tag granularity — settle most of the invoice before any discount contract is signed. The question that shows up in the meeting isn't "how much does it cost?" — it's "what do we optimize first, and with what evidence?".
The problem is that this conversation happens in three dialects. Procurement talks about commitment, term and discount. Engineering talks about invocations, GB-seconds and storage class. Finance talks about cost center and allocation. I was spending too long translating between them in spreadsheets, and every spreadsheet died at the end of the meeting.
What this toolkit solves: putting the four calculations that repeat most — Reserved Instance break-even, S3 class transition, monthly Lambda cost by dimension and the tag schema that supports allocation — in one place, in one language, so the discussion starts at the trade-off and not at the arithmetic.
What it is not: a replacement for Cost Explorer, the Cost and Usage Report or the AWS Pricing API. The numbers that go into the calculators are your assumptions; the tool does the comparison and shows where the tipping point sits. Treat every result as an argument for a conversation, not as an invoice reconciliation.
What's in the box
How it works
The application is a Next.js 16 frontend with React 19, TypeScript 5 and Tailwind CSS 4, living under frontend/. Each calculator is a screen: you enter your assumptions, the screen returns the comparison and the tipping point. There is no backend of its own and no AWS credential involved — which makes the tool safe to open in front of a client or paste into a meeting link without exposing any account.
Pipeline: three GitHub Actions workflows look after the repository — ci (typecheck and build), frontend and security. Deployment goes to Vercel, DNS sits on Cloudflare, and the published version runs at finops.moretes.com. docs/architecture.md describes the structure; OPERATIONS.md and SETUP.md cover operations and environment setup.
What each calculator answers:
- RI vs On-Demand answers "from what utilization does the commitment pay off?". If the instance runs fewer hours per month than the break-even indicates, the discount becomes a loss.
- S3 answers "at what retrieval cost do I switch class?". Glacier and Deep Archive are only cheap for objects nobody reads; the first bulk restore gives the difference back.
- Lambda answers "memory or duration — which dimension do I attack?". Doubling memory usually cuts duration; the calculator shows whether the balance closes.
- Tagging answers "which keys do I require before any resource goes up?". A tag added later is rework across the whole account.
Honest limit: prices are user input. Region, instance type and tier change the unit value — check the Pricing Calculator or the Pricing API before signing anything based on a number generated here.
Flow: from assumption to decision
Everything runs in the browser; the pipeline only publishes the static app. No AWS credential enters the flow.
- RI vs On-Demand · break-even
- S3 class optimizer · Standard · IA · Glacier · Deep Archive
- Lambda estimator · invocações · duração · memória
- Tagging generator · chaves de alocação
- Templates · alocação · showback · chargeback
- ci · frontend · security · typecheck, build, scan
- Vercel · deploy
- Cloudflare · DNS
- Pricing Calculator / Cost Explorer · validação do número
Install and use
- 1
Clone and enter the frontend
git clone https://github.com/fernando-moretes/ref-finops-architect-toolkit.git && cd ref-finops-architect-toolkit/frontend. The whole app lives infrontend/; the root holds docs and workflows. - 2
Install and run
npm installfollowed bynpm run dev. Next.js 16 requires Node.js 20 or newer; ifnpm installfails onengines, that's why. There is no environment variable or AWS credential to configure. - 3
Open the calculator that matches the question
EC2 commitment → RI vs On-Demand. Bucket growing unchecked → S3 optimizer. Function with an odd bill → Lambda estimator. Allocation report that doesn't balance → tagging generator before anything else.
- 4
Validate the number at the source
Take the result to the AWS Pricing Calculator or Cost Explorer with the exact region and type. The calculator points to the tipping point; AWS confirms the unit price of the day.
- 5
Read SETUP.md and OPERATIONS.md before deploying your own
Both files cover environment setup and operations.
CONTRIBUTING.mdexplains how to propose a change; CI runs typecheck, build and security scan on every PR.
git clone https://github.com/fernando-moretes/ref-finops-architect-toolkit.git
cd ref-finops-architect-toolkit/frontend
npm install
npm run dev
# http://localhost:3000
# antes de abrir PR / before opening a PR
npm run lint
npm run buildThe break-even lies when the usage assumption lies
The RI calculator assumes the instance runs the number of hours you entered. In environments I've operated, actual utilization of instances reserved "for production" fell below plan because the team started shutting the environment down at night months after signing. Before committing to a year, take the hours figure from the last 90 days of Cost Explorer — not from the capacity plan.
FAQ
Does the tool read my AWS account?
No. It's a frontend with no backend of its own and no credential. You enter the assumptions; it compares. That's deliberate: you can open it in front of a client with no risk of exposing anything.
Are prices always current?
Treat them as an assumption to confirm. Unit price varies by region, type and tier, and changes over the year. Use the calculator to find the tipping point and the AWS Pricing API or Pricing Calculator for the value of the day.
Why is tagging in a cost toolkit?
Because allocation is the most expensive FinOps problem to fix afterwards. Without CostCenter, Environment and Owner required at creation, the chargeback report doesn't balance and nobody owns the spend. The generator exists so that conversation happens before the first resource goes up.
References
Who it's for
Use this toolkit when you're the architect who has to walk into a cost meeting with a number, not an adjective — and when the question fits one of the four calculations: EC2 commitment, S3 class, Lambda dimension or tag schema. It shortens the distance between engineering, procurement and finance because it speaks the three languages on the same screen. Don't use it as a price source or as an invoice audit: Cost Explorer, CUR and the Pricing API exist for that, and the cost of trusting an unconfirmed number is signing a year of the wrong commitment. If your organization already has a mature FinOps platform, the value here is smaller; if the cost conversation still happens in a disposable spreadsheet, this is exactly the gap it fills.
Architecture, AWS, AI and market deep dives — straight to your inbox. Free.
No spam · unsubscribe anytime