IntSen

Platform · Spend management

Know the cost of every turn—not just the monthly bill.

IntSen costs every governed request and rolls it up to the turn, agent, use case and team—then enforces the budgets and per-turn limits you set.

Budgets in use

Use case · monthly$2,520 of $4,000
Agent · monthly$615 of $1,500
This turn$0.41 of $0.50

Alerts at 50 / 80 / 100 % · deny or alert when exceeded

Illustrative product scenario — synthetic data

The problem

The provider invoice tells you how much. Not who, or why.

A single bill covers every team and agent. A looping agent can burn through budget in minutes, and nobody sees it until the invoice arrives.

  • No attribution to agent, use case or team
  • No limit on a single runaway turn
  • Evaluation and safety-check spend mixed into everything else

Capabilities

What Spend management does

Cost per request

Every governed request is recorded with tokens and cost.

Attribution

Totals per turn, agent, use case and application service, plus per user and per conversation.

Budgets

Monthly budgets for use cases and agents that alert or deny when exceeded; a maximum cost per request.

Per-turn limits

Cost, duration and call limits per turn stop a runaway loop: the gateway refuses the turn's next call.

Threshold alerts

Alerts at 50, 80 and 100 percent of a budget, and usage monitors for unusual activity.

Evaluation spend

Judge and guardrail model calls tagged and shown apart from the agent's own cost.

Technical detail

How cost is attributed and enforced

Budgets are checked in the request path, before the provider is called.

agent.yaml (excerpt)
tasks:  limits: { maxCostUsdPerTask: 0.50, maxDurationSeconds: 900 }budget: { monthlyUsd: 1500, onExceed: alert }
Registration
monthlyUsd, perRequestMaxUsd, onExceed: deny | alert
Agent
monthly budget, plus per-turn limits
Turn
maxCostUsdPerTask, maxDurationSeconds, optional maxRequestsPerTask
Over a turn limit
next call refused with task_budget_exceeded; the turn ends as limit
Alerts
50 / 80 / 100 % of budget; usage monitors for unusual use

Why it matters

For the people accountable for it

FinOps

Attribute AI spend to the agent and team that created it.

AI platform

Set guardrails on spend per use case without blocking innovation.

Engineering

A per-turn limit is the simplest protection against an agent stuck in a tool loop.

Boundaries

What to know before you evaluate

We would rather you hear the limits from us than discover them in a pilot.

  • Cost is calculated from catalog prices and recorded usage; it is an operational figure, not your provider invoice.
  • Chargeback, effective-cost allocation and invoice reconciliation are planned for after the first release.
  • Only governed traffic is costed; spend through direct provider keys is not visible.

FAQ

Common questions

What happens when a budget is exceeded?

With onExceed: deny, further calls are refused with a code. With alert, calls continue and owners are notified.

How do we pick a per-turn limit?

A little above your longest normal turn. The Console shows turn costs and durations to help you set it.

Can we export spend?

Yes. The Console's FinOps view supports CSV export of monthly spend.

Ready to make your enterprise AI accountable?

See how IntSen can help your teams govern access, safety, quality and spending across the AI applications and agents you already run.