Free local AI spend analysis

Find out what the AI bill can—and cannot—tell you.

Analyze a standard CSV export in your browser. See spend concentration, workload attribution, cost per request, period totals, and the measurement gaps blocking a defensible optimization decision.

02 / Bring a usage export

AI Spend Snapshot

Use the standard Snapshot CSV format. Each row can represent one request or an aggregate for a day, model, provider, or workload.

  • Files stay on this device
  • No account or vendor credentials
  • Downloadable aggregate report
  • No unsupported savings promises
CSV format

Required: date, provider, model, cost_usd

Recommended: workload, requests, input_tokens, output_tokens, reasoning_tokens, cached_input_tokens, retries, errors, latency_ms

AI Spend SnapshotLocal analysis

Bring a usage export

Use the standard Snapshot CSV format. Maximum file size: 10 MB.

Download illustrative template

Your file is processed locally and is not uploaded to this site.

What the snapshot produces

A spend baseline and better questions.

The free analysis stays within the evidence in the export. It helps you identify where spend is concentrated and whether the data is ready for a workload-level decision.

01

Concentration

Total spend, cost per request, provider and model mix, workload concentration, and period totals visible in the file.

02

Attribution

The share of spending connected to a named workload and the fields still needed to join cost to production behavior.

03

Field coverage

A transparent, unvalidated file-coverage heuristic and a short list of questions to resolve before changing a model or architecture.

Free baseline / Proprietary depth

Useful on its own. More powerful with the workload.

The public Snapshot tells you what the export supports. A private diagnostic joins spend to traces, quality, pricing, and business outcomes before it recommends a keep, optimize, route, or migrate decision.

Free Snapshot

A private first look

  • Standard CSV import
  • Spend, period totals, and concentration
  • Workload-attribution coverage
  • Transparent field-coverage score
  • Downloadable aggregate report

Runs in the browser. No account, credentials, or file upload.

Diagnostic edition

The decision engine

  • Multi-provider invoice and trace joins
  • Retry, fan-out, and repeated-context analysis
  • Contract and historical pricing
  • Quality-adjusted cost per outcome
  • Keep, optimize, route, or migrate decision

Used privately during a client diagnostic; not exposed as a public SaaS product.

A spend export is not a migration recommendation.

The Snapshot cannot measure task quality, explain trace-level fan-out, apply private contract pricing, or prove that another model will preserve outcomes. Those require production traces, representative evaluations, and business-outcome data.