A deterministic FinOps engine, deliverables an executive can read
The engine computes; consultants and the agent explain, classify, write and converse. No figure is invented: everything comes from your billing, inventory and metrics data, read in read-only mode.
The engine's capabilities, the same for Azure, AWS and GCP
Where the money goes
Cost by service, environment, application, subscription, management group, project or account, in billed and amortized view, over whole calendar months.
Explained variations
From one month to the next, month to date versus the same days of the previous month, from one day to the next: every euro of change is tied to a resource and a cause (new project, growth, prices, optimization, deletion, calendar).
Rules catalogue
Start & Stop, right-sizing confirmed by metrics, disks and snapshots, backup, commitments, managed services, AI costs. One rule, one SQL file, one test; never hidden logic in a connector.
Complete recommendations
Every recommendation carries its evidence, rationale, limitations, prerequisites, risk, commitment impact, owner and status (candidate, published, implemented, verified).
Commitment purchase plan
Reservations, savings plans and CUDs: quantity, family, region, scope and term for the coverage target you choose, with the unused share avoided and the savings plan alternative.
Allocation and chargeback
Business dimensions, virtual tagging (if / then rules) and chargeback rules configured per client; native tags are an input, not the model.
Deliverables and DataViz
Detailed Excel workbook, PDF summary, daily brief, and an interactive DataViz site (French and English, light or dark theme) consuming the same tables, behind password or SSO authentication.
Agent and API
An MCP server and a scheduled agent answer questions on the same data, write summaries and classify anomalies; the agent never invents a figure the engine did not compute.
AI costs
Tracking of OpenAI, Anthropic, Azure OpenAI and LLM gateway spend by team and by client, with re-invoicing: the module dedicated to IT services companies, agencies and software vendors.
The DataViz site on a demo estate
Synthetic "Demo Corp" data: summary, where the money goes, potential by horizon, reservations and Start & Stop, showback, commitment purchase plan, explained variations. Every chart has a table view; light or dark theme; French version.
At our place or yours, always read-only
- No write or administration rights on your resources, ever
- Secrets provided by environment variables or a vault, never stored in deliverables
- Data per client in Parquet files, queried by DuckDB; no database server to operate
- Portal with named accounts and OpenID Connect SSO; strict separation between clients
- Public demos only on synthetic or anonymized data
What the engine reads today
Azure
- Usage details and FOCUS exports (billed and amortized cost)
- Resource Graph: inventory, management groups, tags
- Monitor: VM and disk metrics
- Advisor and reservations
AWS
- Data Exports CUR 2.0 (amortized reservations and savings plans)
- Account and resource inventory
- CloudWatch and Compute Optimizer
- Cost Optimization Hub as a control
GCP
- Detailed BigQuery billing export
- Inventory derived per project
- Recommender and metrics
- Committed use discounts per project
See the engine run on your data
A demo on a real scope is scoped in one call and delivered in a few days, in read-only mode.