Get a Demo

PlatformOperate

Know what every AI dollar bought

Cenna meters model spend the moment it happens — by application, by task, by user, by model. Set budgets and hard caps, approve which models teams can call, and see a spike by name instead of finding it on next month's invoice.

Get a DemoSee the controls
app.cenna.io/platform/costslive · 4 apps
$0+12% vs JuneAI spend · MTD — cloud $992, in your account
Fleet budget0%
$3,400 of $5,000 cap
Token spend by
Contract Hub244.6M tok$1,842
Vendor RiskOver pace88.9M tok$964
HR Onboarding104.7M tok$486
IT Helpdesk36.1M tok$118
Approved models
  • Claude Sonnet 5Allowed
  • Claude Haiku 4.5Allowed
  • GPT-5.2Allowed
  • GPT-5.2 miniAllowed
  • gpt-4oBlocked
AlertBlocked: Vendor Risk attempted gpt-4o — not on the approved model list.

Example data shown for illustration.

01The change

From a monthly surprise to a line item

Before

A surprise invoice at month end

  • One number on a provider bill, weeks after the spend
  • No way to say which app or team drove it
  • Runaway usage only visible once it is too late
  • Every team free to call any model they like

With Cenna

Every dollar named as it is spent

  • Spend attributed by app, task, user, and model, live
  • Budgets and hard caps you set per application
  • Pace alerts that fire before month end, not after
  • One approved-model list, enforced everywhere

02Controls

What you can do with it

The controls that turn AI spend from a monthly surprise into a line item you manage.

  • See spend four ways

    Break the same total down by application, by task, by user, or by model — so "why is this high?" has an answer in seconds.

  • Set budgets and hard caps

    Give each application a budget and a ceiling. When it is reached, spend stops rather than sailing past it.

  • Approve the models

    Choose which models each team may call. A request for anything off the list is refused, not billed.

  • Get warned early

    Pace alerts flag an app trending over budget while there is still a month left to act.

03Model governance

Only the models you approve, ever run

One list, enforced across every application. When an app reaches for a model that is not on it, the call is blocked and surfaced as an alert — so an unapproved model never quietly runs up a bill.

One list · Enforced everywhere · Off-list calls blocked

  • Claude Sonnet 5Allowed
  • Claude Haiku 4.5Allowed
  • GPT-5.2Allowed
  • GPT-5.2 miniAllowed
  • gpt-4oBlocked

04Traceable

Every run is priced on its own

Model spend is not a black box. Each agent run is recorded with what it did and what it cost, so when the number moves you can point at the exact task that moved it.

Every agent run is traced and priced on its own — a clause analyzer at 63 spans and $0.92, a redline drafter at 38 spans and $0.64. When spend moves, you can name the task that moved it.

05FAQ

Common questions

How current is the spend?

It updates as calls happen, not on a daily export. The number you see is the number so far this month, attributed the moment each call returns.

What happens when an app hits its cap?

You choose: a soft cap alerts and keeps running, a hard cap stops further calls. Either way the event is logged, so nothing is silent.

Can spend be split by team or cost centre?

Yes. Applications and their spend roll up to the workspace that owns them, so Legal's usage never lands in IT's budget.

Does this cover every model provider?

Spend is metered across the providers you approve, in one place, so a mix of Anthropic and OpenAI usage shows up on the same screen rather than two bills.

See your AI spend before the invoice does.

Thirty minutes: we break a month of model spend down by app, task, user, and model, and show the caps and alerts that keep it in line.

Get a Demo

Built for speed. Engineered for control.