A GCP cost intelligence platform powered by a custom MCP server — combining a visual spend dashboard with an AI chat assistant for conversational cost analysis, anomaly detection, and month-end forecasting. No data dumps. No pipelines. Live API calls only.
CIRA goes beyond a static dashboard. It combines visual cost intelligence with an embedded AI chat assistant — letting engineers ask billing questions in plain English and get real answers backed by live GCP data.
- 💬 "Why did Prod Dataflow costs spike on Wednesday?"
- 🚨 "Flag anything unusual vs last month"
- 📈 "We're 18 days in — will we exceed budget this month?"
- 🔍 "Which BigQuery datasets are the most expensive this quarter?"
- 📊 "Compare Dev vs QA vs Prod spend this sprint"
No BigQuery exports. No data pipelines. MCP server makes live GCP Billing API calls per request.
| Layer | Technology |
|---|---|
| Cost Data Source | GCP Cloud Billing API (live calls) |
| AI Protocol | Model Context Protocol (MCP) |
| MCP Server | Python (custom built) |
| LLM | Claude API |
| Dashboard + Chat | Streamlit |
| Anomaly Detection | Statistical analysis over API data |
| Forecasting | Time-series projection over billing trends |
Use the embedded chat popup in the Streamlit dashboard. Ask any billing question in natural language — CIRA makes live API calls and returns real answers with numbers.
CIRA automatically compares current spend against historical averages via the Billing API. Flags anything that deviates significantly — by project, by service, or by resource.
Given the current daily burn rate, CIRA projects end-of-month spend and flags whether you're on track or heading for a budget overrun.
- Architecture design
- MCP server tool design
- GCP Billing API integration
- MCP server implementation
- Anomaly detection logic
- Forecasting module
- Streamlit dashboard + embedded chat
- Demo video
Traditional cost dashboards are static — you see what someone decided to show you. MCP turns cost data into a live, queryable AI tool. The MCP server handles all data fetching via live API calls — the Streamlit app consumes it for both the visual dashboard and the chat popup. One source of truth, one unified interface.
Krishna Ramadas — Senior Data Engineer 4+ years of GCP and Snowflake cost optimisation experience, including $250K+ in documented savings for a Fortune 500 client.
MIT License — see LICENSE for details.