What I Built
In a sprawling cloud environment, "zombie" resources specifically unattached EBS volumes are a silent drain on the budget. As a Cloud Architect, I wanted to build a tool that doesn't just manage infrastructure but optimizes its cost.
I built Cost-Sentry, a FinOps agent that identifies unattached EBS volumes and calculates their financial impact. It uses the GitHub Copilot CLI as a reasoning engine to generate complex Boto3 auditing scripts, which are then used to produce immediate, actionable saving reports.
Demo
The tool is designed for high-velocity cost audits and is now part of my public GitHub portfolio.
🔗 GitHub Repository: https://github.com/mpawar006/cost-sentry
Cost-Sentry in Action
The tool scans for volumes in the available state and applies a standard rate of $0.10 per GB/month to estimate waste.
Figure 1: Cost-Sentry successfully identifying 20GB of unattached storage waste.
Financial Impact Detected:
Total Volumes Audited: 2
Total Wasted Capacity: 20 GB
Potential Monthly Savings: $2.00
My Experience with GitHub Copilot CLI
Building this fourth project on my local machine solidified how AI can be a force multiplier for cloud management.
Logic-Driven Prompt Engineering: Unlike standard code-gen, I used a structured cost_library.json to feed specific financial archetypes into the GitHub Copilot CLI. This allowed the AI to focus on the mathematical reasoning required for cost estimation rather than just boilerplate code.
The "Safety-First" Approach: A major takeaway from this challenge was ensuring AI-driven velocity doesn't bypass safety guardrails. I engineered Cost-Sentry as a read-only auditor, proving that AI agents can be highly impactful without needing high-risk permissions.
FinOps Visibility: The CLI allowed me to quickly prototype the logic needed to parse deeply nested AWS resource metadata and transform it into a clean, human-readable terminal table using the tabulate library.