What are NTUs?

What are NTUs?

Business Summary: NTUs give finance and ops leaders a predictable unit for AI work spend — a quoted cost before each run and monthly pools for budgeting, instead of opaque token bills.

An NTU (Nimbus Token Unit) is Nimbus's customer-facing unit for measuring governed agent work — not raw LLM tokens, but a normalized estimate of compute, orchestration, and connector activity for a run.

Before you commit to a Plan mode run, you need predictable economics: what will this run likely cost, will it fit our monthly budget, and can we compare agents and workstreams fairly? NTUs answer those questions with a single number on the Confirm & run card and in billing reports.

How this fits in Nimbus

NTUs connect agent execution to billing and governance. Every governed Plan run shows a quote before you confirm; usage posts to your monthly pool and appears on the Usage & Billing dashboard.

flowchart LR
  Plan[Plan run proposed] --> Quote[Quote and ceiling]
  Quote --> Confirm[Confirm and run]
  Confirm --> Execute[Run executes]
  Execute --> Meter[NTU metered]
  Meter --> Pool[Deduct from monthly pool]
  Pool --> Dash[Usage dashboard]
  Pool --> Ovg[Overage if applicable]

Why NTUs exist

NTUs normalize agent work into one billing-friendly number. Finance can set monthly pools by plan tier, operators see quotes before approving runs, and governance can enforce spend thresholds at HITL checkpoints — all in the same unit.

Quote and ceiling

Every governed run shows:

Field Meaning
Estimated NTU (quote) Expected cost for the planned run
Ceiling Upper guardrail — typically around 1.5× the quote

Nimbus checks quote and ceiling against your included NTU pool and any run budget limits before execution starts. If a run would exceed limits, confirmation is blocked with an explanation.

See Run budget limits.

Monthly pool

Your subscription includes a fixed NTU pool each billing period:

  • Pro — 200 NTU
  • Team — 2,000 NTU
  • Business — 5,000 NTU

Optional usage commitment add-ons increase the pool on Team and Business. See Plans, seats, and included usage.

Usage accumulates as runs complete. The header meter and Usage & Billing tab show progress through the period.

What consumes NTUs

Primarily Plan mode governed agent runs — multi-agent orchestration, report synthesis, and approved connector reads/writes.

Casual Perception chat and lightweight Analyze interactions may consume fewer or no metered NTUs depending on the governance tier your admin has set — see Org access tiers; formal governed runs are the main metered path.

See .

NTUs vs decision commits

NTUs measure execution cost. A decision commit is an approved outcome — publishing a report or pushing to a CRM. You can approve a decision after NTUs were consumed during the run that produced it.

See What is a commitment?.

Ready to try it?

Example scenario

Your Team plan includes 2,000 NTUs per month. A RevOps lead scopes a pipeline analysis workstream and sees Estimated NTU: 38 (ceiling 57) on Confirm & run. After sign-off on the output, 41 NTUs post to the ledger — slightly above the quote but under the ceiling. The header meter moves from 1,200 to 1,241 used. Finance checks the Usage tab weekly; no overage yet because the team stays under 2,000 included.

FAQs {#faqs}

Q: Is one NTU one API token? A: No. NTUs normalize agent work into a billing-friendly unit. Internal token usage is abstracted away.

Q: Do failed runs still charge NTUs? A: Partial usage may post depending on how far the run progressed. Check the usage ledger entry status on your dashboard.

Q: Can I see NTU economics before assigning an agent? A: Exact quotes appear on the Confirm & run card before a workstream run starts.

Q: Who should monitor NTU usage? A: Workspace admins and finance partners watch the Usage tab; contributors see quotes before each run they confirm.

Q: Where do NTUs show in the UI? A: See Where NTUs appear.

Q: Do NTUs reset each billing period? A: Yes. Included pool resets each period; unused NTUs typically do not roll over. Confirm on your contract.