Podmonkey 🐒

Paste Kubernetes YAML. Compare cloud bills. Spot waste before you deploy.

Podmonkey is an open-source, manifest-first Kubernetes cost estimator. It parses your YAML and produces planning-grade monthly cost ranges across AWS EKS, Google GKE, Azure AKS, and Hetzner — no cluster agent, no login, no enterprise sales call.

Planning estimates only — not an invoice. See docs/CALCULATION_PLAN.md for how we make numbers decision-grade, and docs/CURRENT_STATE.md for what the code does today.

Quick links

| Doc | Description | |-----|-------------| | Current state | Audit: claims vs code, gaps, extension points | | Calculation plan | Dual models, rate derivation, validation | | Product spec | Vision, scope, roadmap, warnings | | Methodology | Formulas, price sheet schema | | Competitors | Side-by-side vs Kubecost, ReleaseRun, Optiqor, etc. | | GitHub Action | PR cost comments, diff, policy gates in CI | | Roadmap | Tier 2–4 feature plan | | Pricing sources | Verified list prices and benchmarks |

Try it

Live demo → — paste YAML, compare AWS / GCP / Azure / Hetzner in one view. Load examples from the dropdown (nginx, Postgres HA, SaaS stack, monitoring, and more in examples/).

Or run locally:

cd apps/web
npm install
npm run dev
# → http://localhost:3002

Deploy your own copy: Vercel (set root to apps/web) or enable GitHub Pages in repo settings (workflow .github/workflows/deploy-web.yml publishes on push to main).

Add to your repo

Copy .github/workflows/podmonkey-estimate.example.yml to .github/workflows/podmonkey.yml and point path at your manifests:

- uses: MarketMadi/[email protected]
  with:
    path: ./k8s
    base-path: ./k8s   # optional: compare vs main branch copy for PR diff
    max-monthly-usd: '500'
    min-confidence: '60'

Podmonkey posts (or updates) a PR comment with per-provider monthly ranges and optional cost diff. See docs/GITHUB_ACTION.md.

Install (CLI)

npm install -g podmonkey
# or without installing:
npx podmonkey estimate -f examples/nginx-deployment.yaml

Helm charts:

podmonkey estimate --helm-chart ./chart --helm-values values.yaml
helm template myapp ./chart | podmonkey estimate -f -

PR cost diff:

podmonkey estimate -f k8s/ --base k8s.main/ --markdown

How it works

YAML manifests
    → parse (Deployment, StatefulSet, PVC, Service…)
    → aggregate CPU/RAM/storage/LB counts
    → derive rates from reference_instance (OpenCost Appendix A)
    → compute marginal .. node-floor range per provider
    → compare providers + confidence score + warnings

Compute (two models, shown as a range):

Model A (marginal):  Σ(requests) × 730h × derived $/vcpu-hr and $/gib-hr
Model B (node floor): cheapest catalog VM that fits workload × hourly × 730h

Plus control plane, PVC storage ($/GiB-month), and LoadBalancer flat fees per provider.

Rates are derived at runtime from each sheet's reference_instance — not the legacy rates fields in JSON.

Example (nginx on AWS EKS): ~$103/mo without a LoadBalancer; ~$121/mo with one (on-demand, planning estimate).

Project structure

podmonkey/
├── docs/           # Product, methodology, audit
├── pricing/        # Versioned provider price sheets (JSON)
├── src/
│   ├── cli/        # CLI entry (estimate command)
│   ├── parser/     # YAML → normalized workloads
│   ├── pricing/    # Rate derivation, node floor, confidence
│   ├── estimator/  # Cost engine
│   └── warnings/   # Policy rules
└── apps/web/       # Next.js demo UI

Development

# Root: shared engine tests (41 tests)
npm install
npm test
npm run build

# CLI (also: npm install -g podmonkey)
npm run podmonkey -- estimate -f examples/nginx-deployment.yaml
npm run podmonkey -- estimate -f k8s/ --base k8s.main/ --max-monthly-increase-usd 50
npx podmonkey estimate -f examples/nginx-deployment.yaml
kubectl get deploy,svc -o yaml | npx podmonkey estimate -f - --json

# GitHub Action (PR comments + policy gates) — see docs/GITHUB_ACTION.md
# uses: MarketMadi/[email protected]

# Web app
cd apps/web && npm install && npm run dev

Pricing data

Price sheets live in pricing/ with as_of dates, reference_instance, and source URLs:

• pricing/aws-us-east-1.json
• pricing/gcp-us-central1.json
• pricing/azure-eastus.json
• pricing/hetzner-fsn1.json (compute_model: node_only)

Rate derivation: docs/CALCULATION_PLAN.md. CI asserts every sheet normalizes to its reference VM hourly price.

Differentiation

| vs Kubecost | vs ReleaseRun | vs DeepCost | |-------------|---------------|-------------| | No cluster install | Prices your YAML, not sliders | Open source, not SaaS funnel | | Multi-cloud compare in one view | Per-workload breakdown | Transparent methodology | | Honest range (marginal .. node floor) | — | — |

License

MIT — see LICENSE.

Disclaimer

Podmonkey is not affiliated with AWS, Google, Microsoft, Hetzner, Kubecost, or OpenCost. Estimates use public on-demand pricing and resource requests from your manifests. Excludes egress, NAT, Spot/reserved discounts, and idle node capacity. Actual bills depend on usage, discounts, networking, and bin-packing.