Comparable or not?
See when monetary reconciliation is supported and when the available source fields make it NOT_COMPARABLE.
US-first · fixed scope · one-time engagement
Turn scattered LLM exports into evidence you can inspect.
In 48 hours, we map the OpenAI and Anthropic billing and usage exports you already have into a source-linked workbook, coverage record and bounded next-check plan.
No API keys. No production access. No subscription. Do not attach exports to the fit-check email.
Two providersOpenAI and Anthropic
Defined input31 days · 50k rows · 50 MB
Traceable outputWorkbook, mappings and proof manifest
Protected intakeNo credentials or production access
The problem
You have dashboards. You still need evidence that joins up.
Provider cost reports, usage reports and internal allocation records often expose different levels of detail. This diagnostic makes those boundaries explicit instead of forcing unlike values to agree.
Coverage and mappings
See exactly which provider, model, project and workspace fields were mapped, absent or unsupported.
Evidence-backed checks
Surface exact duplicates and unmapped dimensions, plus retry or context observations only when explicit fields support them.
Exact scope
Small enough to finish. Detailed enough to act on.
Scope is confirmed before files move. The delivery clock begins only after VANTYX Labs confirms DATA ACCEPTED in writing.
Documented native API JSON or the published cost-bearing normalized schema, accepted as CSV or a JSON array.
One agreed time window, using the source timestamps supplied.
Combined maximum across complete accepted CSV and JSON source files.
Starts only after written DATA ACCEPTED, never at checkout or first contact.
An active, bounded reconciliation question.
- You use both OpenAI and Anthropic.
- You can export authorized billing and usage records.
- You have one defined 31-day period to investigate.
- An accountable person can clarify source labels and dates.
Clear boundaries protect the result.
- Live account or credential access
- More than two providers or 50,000 rows
- Raw prompts, customer content or unrelated personal data
- Continuous monitoring or remediation implementation
- Legal, tax or financial-statement assurance
- Guaranteed savings or charge recovery
What you receive
A complete diagnostic package, not a mystery spreadsheet.
Record-level findings retain source references; summary totals remain tied to accepted hashed files. Unknowns and unavailable comparisons stay visible instead of being filled with assumptions.
Coverage manifest
Accepted files, SHA-256 hashes, source schemas, mappings, limits and explicit capabilities.
Normalized dataset
A consistent provider, model, timestamp, usage and cost structure with record-level lineage.
Reconciliation workbook
Coverage, source mapping, supported totals, differences, exceptions and bounded checks in one workbook.
Variance register
Supported provider variances, exact duplicates and unmapped dimensions with source references and limitations.
Executive report
A concise statement of what was supported, what was not comparable and why.
Cost-driver evidence
Provider/model concentration plus retry and context observations only where source fields support them.
Method and action plan
Deterministic rules and up to five source-linked verification actions, with no invented savings estimate.
Written findings handoff
One structured email handoff with the supported findings and prioritized next checks.
DEMONSTRATION ONLY · SYNTHETIC DATA · NOT A CUSTOMER RESULT
Run-generated proof, not hand-written claims.
The public sample is generated by the diagnostic engine from a deterministic synthetic fixture. It proves the output format and arithmetic—not customer savings, demand or past performance.
diag_108c3507db7c171b256d2b24Deterministic synthetic run
Usage coverage6 received · 5 canonical
Billing coverage2 received · 2 canonical
Arithmetic$150.00 usage · $154.00 billed
Observed difference+$4.00 billing minus usage
Customer resultNone — synthetic proof only
Two supported evidence paths
The source determines what can be concluded.
The engine does not silently price tokens or pretend that provider-native usage records contain request-level dollar cost when they do not.
| Accepted input path | Cost total | Usage drivers | Monetary reconciliation |
|---|---|---|---|
| Provider-native JSON Documented OpenAI and Anthropic usage/cost responses, complete pages only | Provider-reported only Not independently verified against an invoice; source categories not present in the report remain outside the result. | Supported where exported | NOT_COMPARABLE Usage lacks explicit cost; no independent repricing is performed. |
| Cost-bearing normalized schema Published CSV headers; the identical keys are accepted in a JSON array, with explicit source-supplied USD costs | Customer-supplied billing total | Supported where supplied | SUPPORTED Exact arithmetic on supplied values; those values are not independently repriced. |
Both paths retain source mappings, SHA-256 lineage, coverage limitations and deterministic output files. See the full methodology.
How it works
Fit first. Files only after a secure channel is agreed.
This is a diagnostic service, not software access. The fit-check contains no exports or secrets.
Check fit
Email only the scope, providers, period, format and approximate size. Do not attach files.
Confirm scope
Receive written confirmation that the request fits the fixed engagement.
Purchase
Use the applicable one-time Stripe checkout after scope is clear.
Secure intake
Agree the protected transfer channel and data-handling terms before any export moves.
DATA ACCEPTED
After readability and scope checks pass, written acceptance starts the 48-hour window.
Data boundaries
Minimum data. Explicit acceptance. No account access.
Fit is established without files. The transfer channel, processing location and retention arrangement are agreed in writing before transfer.
Never requestedAPI keys, passwords, production access, raw prompts or customer content.
Accepted only after purchaseThe minimum authorized CSV or JSON exports required for the agreed scope.
RetentionNo generic deadline is claimed. The exact retention and deletion arrangement is agreed in writing before transfer.
Policy detailRead the data-handling policy and privacy notice.
Transparent pricing
One scope. Two time-bound price points.
The seven-day launch rate changes only the price, not the scope or delivery standard. Stripe processes payment; there is no subscription or automatic renewal.
$149 USD
Available from through .
Launch opens soon — $149$500 USD
Applies after the seven-day launch period for the same fixed scope.
Available after launch — $500Check fit before paying. Sold by Diogo Ribeiro, trading under the VANTYX Labs commercial brand in Portugal. Full seller details and refund conditions are stated in the service terms.
FAQ
Know the boundaries before you begin.
Questions not answered here belong in the no-file fit-check.
Which providers and formats are supported?
Exactly OpenAI and Anthropic over one exact shared period. The engine accepts documented native usage/cost API JSON responses as complete pages and the published cost-bearing normalized schema as CSV or a JSON array with the identical keys. Normalized billing requires one undimensioned provider-total row per provider; native cost buckets must be exact-overlap components or contiguous intervals without gaps or partial overlaps. During fit-check we inspect field names only. The combined limit is 50,000 rows and 50 MB for a period up to 31 days.
Can native provider JSON produce an exact dollar reconciliation?
Not by itself. Native cost reports provide provider-reported cost totals and native usage reports provide usage drivers, but those totals are not independently verified against an invoice and may omit categories not present in the source report. They are marked NOT_COMPARABLE when usage records do not contain explicit cost. We do not silently reprice tokens. Exact arithmetic requires the cost-bearing normalized template, and even then source-supplied costs are not independently repriced.
When does the 48-hour clock start?
Only after purchase, secure transfer, readability and scope checks, and written confirmation from VANTYX Labs stating DATA ACCEPTED. Checkout or first contact does not start the clock.
What must never be shared?
Do not share API keys, passwords, production credentials, wallet data, raw prompts, customer content or unrelated personal data.
How are files transferred and retained?
The secure transfer channel, processing location and exact retention/deletion arrangement are agreed in writing before any export moves. Files must not be attached to the fit-check email.
What does the engine actually flag?
It reports supported provider variances, exact duplicate records and unmapped model/project/workspace dimensions. Retry and context-growth observations appear only where every required source field is present. No observation is automatically labeled waste, overbilling, savings or causation.
What if the totals do not fully reconcile?
Supported differences remain explicit in the variance register with source references, limitations and recommended next checks. Unsupported monetary comparisons remain NOT_COMPARABLE; missing values are never invented.
Does the diagnostic guarantee savings or recovered charges?
No. It provides a traceable reconciliation and prioritized next checks. It does not guarantee savings, refunds from providers, remediation or a particular business outcome.
Is this an accounting or legal audit?
No. It is a technical fixed-scope diagnostic of supplied LLM billing and usage exports, not legal, tax, accounting or financial-statement assurance.
Can I add providers, periods or more data?
Not within this fixed scope. Use the fit-check to describe a different requirement before purchasing; no expanded scope is implied by this offer.
Start without sharing files
Check fit by email
Confirm fit before you pay.
Send only the providers, period, format, approximate size and problem to resolve. Never attach exports or credentials to the first email.