caipe-io AWS Cost Analysis
Query AWS Cost Explorer data to break down spending by service, identify trends, detect anomalies, and recommend cost optimization actions.
Instructions
Phase 1: Cost Breakdown
- Fetch cost data using the AWS agent's Cost Explorer tools:
- Total spend for the requested period (default: last 30 days)
- Breakdown by AWS service (EC2, EKS, S3, RDS, Lambda, etc.)
- Breakdown by account (if multi-account)
- Daily/weekly cost trends
- Calculate key metrics:
- Month-over-month change (absolute and percentage)
- Top 5 most expensive services
- Top 5 fastest-growing services (by percentage increase)
Phase 2: Anomaly Detection
- Identify cost anomalies:
- Services with >20% cost increase vs. previous period
- Sudden spikes in daily spend
- New services that appeared this period
- Unused resources still incurring charges
Phase 3: Optimization Recommendations
Provide actionable recommendations with estimated savings:
- Reserved Instances or Savings Plans opportunities
- Right-sizing overprovisioned resources
- Storage tier optimization (S3 lifecycle policies)
- Spot Instance candidates for non-critical workloads
Output Format
```markdown
AWS Cost Analysis Report
Period: Feb 1 - Feb 9, 2026 | Total Spend: $24,567.89
Top 5 Services by Cost
| Service | Cost | % of Total | MoM Change |
|---|---|---|---|
| Amazon EKS | $8,234 | 33.5% | +12% |
| Amazon EC2 | $6,789 | 27.6% | -3% |
Optimization Opportunities
| Recommendation | Estimated Monthly Savings |
|---|---|
| Right-size 3 oversized EC2 instances | $450/mo |
| Purchase EKS Savings Plan | $1,200/mo |
| ``` |
Examples
- "Show me the AWS cost breakdown for the last month"
- "What are our top 5 most expensive AWS services?"
- "Are there any cost anomalies this month?"
- "How can we reduce our AWS spend?"
Guidelines
- Default to last 30 days if no time period is specified
- Always show both absolute costs and percentage of total
- Include month-over-month comparison for context
- Round costs to 2 decimal places for totals, whole dollars for individual services
- Be specific in optimization recommendations (name the resource, not just the category)