AI spend

Your AI bill, under control, provider by provider

OpenAI, Anthropic, GitHub Copilot, Cursor and OpenRouter in one panel, next to your cloud. See spend by provider and model, rules flag what is avoidable and, on Business, the agent sets caps, alerts or revokes keys with your approval.

  • 5

    AI providers you can connect

  • 16

    rules that flag avoidable spend

  • With your OK

    caps, alerts and revocations

  • AI + cloud

    in the same panel as AWS, Azure and GCP

The problem

Why the AI bill gets away from you

Every provider, its own console

OpenAI, Anthropic, Copilot and Cursor are paid separately. Nobody sees the total or who each part belongs to.

Caps and alerts nobody sets

A key with no limit, an unused seat or someone who left keep spending until the bill arrives.

The expensive model by habit

Simple tasks on top-tier models, without prompt caching or at standard price when they could run as batches.

Every provider on its own, and all of it in one bar

  • OpenAI
  • Anthropic
  • GitHub Copilot
  • Cursor
  • OpenRouter
The solution

How AAA Sentinel solves it

  1. Connect your providers

    OpenAI or Anthropic included on Pro (the other one at extra cost); GitHub Copilot, Cursor and OpenRouter too on Business.

  2. See the spend in one place

    By provider and by model, in the same panel as your AWS, Azure and GCP cost.

  3. Rules flag what is avoidable

    On Pro, 5 efficiency rules: prompt caching, batch-ready load, context window, expensive model for simple tasks and cache writes. On Business, 11 more, including: no limit or no alert, unused keys, idle seats, people who left, sudden spikes, unsupervised automated spend and spend near the limit.

  4. Act with your approval

    On Business, the agent sets caps and alerts or revokes keys and seats, as far as each provider allows. Nothing is applied without your approval.

At a glance

What the agent can do on each provider

It depends on what each provider allows through its API. On Business and always with your approval.

What the agent can do on each provider
ProviderSpend capSpend alertRevoke keyRevoke seatPer-user cap
OpenAIYesYesYesNoNo
AnthropicNoNoYesNoNo
GitHub CopilotYesYesNoYesNo
CursorNoNoNoNoYes
OpenRouterYesNoYesNoNo

Anthropic does not expose spend caps or alerts in its admin API for Claude Console organizations, so the agent can only revoke keys. On Copilot, the cap covers metered usage, not seats. Cursor鈥檚 per-user cap requires its Enterprise plan.

What each plan includes

Pro

  • OpenAI or Anthropic (the other one at extra cost)
  • Spend by provider and model
  • 5 efficiency rules

Business

  • Plus GitHub Copilot, Cursor and OpenRouter
  • 11 governance and anomaly rules
  • Agent caps, alerts and revocations, with your approval
See plans

What it does not do

  • On Anthropic (Claude Console organizations) it cannot set caps or alerts: its admin API does not allow it. It only revokes keys.
  • On Pro there are no agent actions: you see the spend and the efficiency rules, and you act.
  • On your AI providers it never applies a change without your approval.

Frequently asked questions

From Pro, OpenAI or Anthropic (the second one at extra cost); on Business, GitHub Copilot, Cursor and OpenRouter too.