Day 25 — DevOps
“Why is our AWS bill $4,200 this month?” is a question that ruins mornings.
An MCP server that tracks cloud spend across providers, flags anomalies, forecasts next month, and tells you exactly where the money is going. No more spreadsheet archaeology.
Example Queries
Tools
| Tool | What It Does |
|---|---|
| cost_breakdown | Breakdown of cloud costs by provider, category, or date range — total spend, daily averages, and per-service totals. |
| service_costs | Drill into a specific service — daily cost history, usage metrics, trends, and month-over-month changes. |
| budget_status | Budget status across all providers — actual vs. budget, burn rate, projected overage, and month-over-month comparison. |
| anomaly_detection | Detected cost anomalies — spikes, unusual patterns, and unresolved issues with severity and root cause notes. |
| cost_forecast | Cost forecasts for the next 1–3 months — predicted spend per provider with confidence intervals and budget comparison. |
| provider_comparison | Compare cloud providers side by side — total spend, cost per category, budget utilization, anomaly count, and trends. |
The Pipeline
Every dollar is tracked across providers, services, and categories — then analyzed for patterns and problems:
Daily cost records from every service across AWS, GCP, Azure, and Vercel — with usage quantities and units for context.
Break down spend by provider, service, and category. Spot which services are trending up and which are stable.
Detect cost anomalies automatically — spikes from misconfigured auto-scaling, runaway queries, or unexpected API volume.
Project next month's bill with confidence intervals. Know if you're headed over budget before it happens.
Why This Build
Cloud costs are the new electricity bill — everybody has one, nobody knows exactly what's in it, and it keeps going up. AWS, GCP, and Azure billing dashboards are powerful but siloed. Nobody has time to check three different consoles every morning.
The real problem isn't the total number — it's the surprises. An unpartitioned BigQuery scan that costs $112 instead of $40. An ECS deploy that doubled your task count because of a health check bug. A new AI feature that tripled your API calls overnight.
This tool catches those surprises before they compound. Ask one question, get a cross-provider view of where the money is going, what's abnormal, and what next month looks like. Second DevOps build of the challenge.
Demo Data
4
Providers
21
Services
1,890
Cost Records
12
Budget Entries
7
Anomalies
90
Days of Data
4 providers (AWS, GCP, Azure, Vercel) with 21 services across compute, database, storage, network, analytics, AI, and messaging categories. 90 days of daily cost records with realistic trends — some services stable, others trending up. 7 anomalies including a $112 BigQuery scan, a recursive Lambda, and a doubled ECS deployment. Forecasts for the next 3 months with confidence intervals.
Setup
cd day25-infrastructure-cost
npm install && npm run seed && npm run buildThen add the config to Claude Desktop and restart.
If you're checking three billing dashboards every morning instead of asking one question, let's fix that.