Skip to main content

​
nOps AI

AI is where nOps brings your AI spend together: model and token usage across providers, per-seat spend and limits, the requests people file when they hit a cap, and the budgets and wallets that decide who gets what. It sits on the same Explorer and Budgets machinery as the rest of nOps, so AI spend is allocated and budgeted the same way your cloud spend is.
AI requires the Inform plan. Some areas need additional permissions — see Who sees what below.

​
Get started

  1. Sign in to nOps.
  2. Connect at least one AI provider integration (Claude Enterprise, Cursor, or ChatGPT / Codex) in Settings → Integrations.
  3. In the sidebar, open AI → Tracking to see model and token usage right away, at clara.nops.io/ai-optimization/tracking.
  4. To manage seats, limits, and budgets, create a seat tracking project. See Set up AI seat tracking.

​
What’s in the AI area

Tracking

Model Overview and AI Explorer — which models and platforms your spend and tokens go to.

Requests

The queue where people ask for a higher spend limit and reviewers approve, reject, or ask for more.

Wallets

Split AI spend into groups, assign seats to them, and pick who approves each group’s requests.

Budgets

Per-group and per-seat targets, plus a Budgets vs. Actual forecast view.

Recommendations

Measured model, seat, and cap findings across Bedrock, Cursor, and Claude Enterprise.

Anomalies

AWS-detected anomalies filtered to AI services, plus Clara’s analysis.
One more surface sits alongside these:
  • Automated Rebalance — an admin-only tab inside seat tracking that moves unused limit to the seats about to hit their cap.

​
Key concepts

​
Seat tracking projects

Most of the AI area is organized around a seat tracking project. A project ties together three things that already exist elsewhere in nOps: an AI-scoped Cost Allocation showback (your wallets), a budget with an Explorer report behind it, and the live seat rosters your provider integrations sync. Creating the project provisions those links for you. Tracking works without a project. Wallets, Budgets, Automated Rebalance, and most of the Requests workflow do not.

​
Providers

nOps tracks three AI providers today, and they do not all do the same things: Only Claude Enterprise can have its per-seat spend limits read and written by nOps, which is why Automated Rebalance and applied limit changes are scoped to it. Cursor and ChatGPT / Codex spend is still tracked, allocated, budgeted, and forecast — it is simply read-only. Model and token usage on Tracking covers a wider set of platforms than the seat providers, including Amazon Bedrock spend that arrives through your cloud integration.

​
Who sees what

The AI group in the sidebar adapts to your permissions:
  • Everyone with AI access sees Tracking, Requests, Recommendations, and Anomalies.
  • Seat tracking admins additionally see Wallets and Budgets, and get the admin-only Automated Rebalance tab inside seat tracking.
  • A restricted request-viewer role sees only Requests, and only their own.
Model Overview and AI Explorer are each gated independently for your organization. If only one is enabled, the Tracking page shows it directly with no sub-tabs.

​
FAQs

No. AI → Tracking works as soon as an AI provider or cloud integration is sending spend. A project is what unlocks wallets, budgets, seat limits, and requests.
Those two sidebar entries require seat tracking admin rights. Without them you still see Tracking, Requests, Recommendations, and Anomalies.
Only for Claude Enterprise, only when an admin has switched the workflow off Observe, and only within the rules described in Automated Rebalance. Cursor and ChatGPT / Codex limits are read-only in nOps.
A seat tracking project creates a Cost Allocation showback and an Explorer report behind the scenes, and its budget is an ordinary nOps budget. You can open any of them from their own pages and they behave exactly as documented there.