Questions leads ask before they install.

Partly, and some of those dashboards are very good. Anthropic and Cursor each report cost per person13, OpenAI reports billed amounts by project, line item or API key2, and the Copilot usage metrics we read carry no cost at all4.

That is four cost models that do not line up, and each vendor dashboard is built around its own tools. ai-cur puts them on one page, in one currency, beside the same people.

LiteLLM is a strong gateway: spend per key, user and team, routing, caching and hard budget limits, self-hosted, with no licence price5. Running it still costs you operations time.

A gateway sees the traffic that passes through it. Seat invoices reach you by email, not through a proxy, and neither do your merged pull requests. ai-cur reads both, and serves a cap decision that LiteLLM or any gateway can enforce.

Several do, and some at token level. What we did not find in the public documentation we reviewed is that impact reconciled to the invoice, with seats, credits and discounts in it.

That is absence of evidence rather than proof, so ask your vendor where credits and discounts appear. If they can show you, you may not need us for that part.

It may well see individual developers. Vantage breaks Cursor spend down by developer6, and CloudZero splits Anthropic Enterprise spend by user, separating Claude Code from chat7.

A FinOps tool is not built to hold what shipped: commits, pull requests, changed lines and projects. That half, set beside the spend, is what ai-cur adds.

They will build more. Gartner's 2026 evaluation of AI coding agent vendors counts cost controls among the capabilities it expects8, and vendor analytics already come with the plan.

The most capable platform budget we reviewed, Unity AI Gateway on Databricks, leaves provisioned throughput and inference on outside models untracked9. A platform's controls end where spend on other vendors begins, and that is where your waiting would leave the gap.

We have no accuracy figure to give you. None of the vendors we reviewed publishes one either, and a number without a method would be worse than none.

ai-cur makes no attribution claim: it sets a person's settled spend beside the same person's merged work and never says a tool wrote it. If we ever publish attribution, we will publish how it was measured, how wrong it can be, and where it is blind.

No. The record is aligned with FOCUS version 1.4, ratified on 4 June 202610, and keeps model and input and output tokens apart where the vendor reports them.

It has not been through a validator, and a FOCUS certification needs Foundation membership and credentials we do not hold. Until then, alignment is the whole claim.

We will not name one. Anyone who quotes a percentage before reading your bill is guessing.

What you get is the bill laid out so you can find the idle seats, the overruns and the plans that no longer fit. Any saving is yours to measure, in your own records.

That is a fair worry about incentives. A price that grows with your spend would reward us when you spend more, and a cut of savings would make every saving an argument.

So Team is priced per managed seat and Business per contract or by a platform fee. No share of your spend is on the price list.