Skip to content

Use casesSpend control

A spend limit that refuses the request, not one that reports it.

Set a ceiling per organization, project, member or gateway key. A request the limit cannot cover is refused with a 402 before your provider sees it, so a loop, a leaked key or a busy agent stops at the number you chose.

Start 14-day trialRequest a demo
Key ci-agent · limit $50.00 / month · blockingExample
Spent so far$49.82 estimated provider cost
Next requestestimated $0.31
Result402 budget_exceeded · code key_budget
Providernot called

Example key. The limit counts INFRO's estimate of provider cost, not the provider's invoice.

The problem

Without a control plane

  • A provider dashboard tells you what was spent, after it was spent.
  • One runaway script or leaked key can use a month's budget in an afternoon.
  • Every team shares one provider key, so nobody owns a number.

With INFRO

  • Limits are checked before the provider call, on every request.
  • Each project, member and key has its own ceiling and alert thresholds.
  • A refusal names the scope that refused it, so the right person acts.

How it works

  1. Give each workload its own key

    A key per service, agent or developer, each inside a project. Spend is recorded per key from the first request.

  2. Set the ceilings

    Limits at organization, project, member and key level, each blocking or alert-only, with alerts at your own thresholds.

  3. Let the gateway enforce them

    The estimated cost of each request is reserved before the call. A request the limit cannot cover returns 402 budget_exceeded and never reaches the provider.

What you get

Each one is documented, and each says which plan includes it. Where no plan is named, it is on every plan.

  • Limits at every level

    A hard cap per organization, project, member and key, with alerts at your own thresholds.

    Docs

  • Refused before the call

    A request a limit cannot cover is refused with 402 before the provider sees it; the error names the scope.

    Docs

  • Spend alerts and forecasts

    A day that departs from its usual pattern, or a budget forecast to be exceeded, becomes a finding by email, Slack or webhook.

    Docs

  • Key limits

    Limit a key to chosen providers or IP ranges, turn it off, or rotate it with an overlap window.

    Docs

  • Caps per end customerFrom Team

    One spend cap for every end customer of your product, with overrides your backend sets.

    Docs

  • Unpriced traffic refused

    A model with no known rate is refused under a limit unless you allow unpriced traffic, so a limit never counts a guess as $0.

    Docs

Who it is for

  • Engineering leadsGive every developer and agent a key with a ceiling, and stop approving spend by trust.
  • FinanceA number per team that is enforced, not a number per team that is hoped for.
  • FoundersA runaway loop costs at most the limit you set on that key.

FAQ

Is the limit exact to the cent of my provider's invoice?

The limit counts INFRO's estimate of provider cost from the rates it knows, or your negotiated rates where you entered them. The provider's invoice is the final figure; reconciliation compares the two where the provider publishes cost data.

Does a limit hold under concurrent requests?

Each request reserves its estimate before the call, so a burst cannot all pass a check at once. The method and the results of a sandbox test at 500 concurrent requests are published in the field guide linked below.

Can a limit alert instead of block?

Yes. Each limit is blocking or alert-only, and alert thresholds are yours to set.

The full contract is in the Spend controls docs, and every other question is on the FAQ.

Field guides

Related

All use cases

Get started

Try it on your own provider accounts.

Connect a provider, create a gateway key and change one base URL. Your providers keep billing you directly; 0% markup on model spend.

Start 14-day trialRequest a demo