Usage

See how much model work your workspace did, by agent and by model, find the heaviest users, and export the report.

For owners and adminsRun your workspace · page 14 of 35

Usage shows how much model work your workspace did over today, the last 7 days or the last 30 days. It counts tokens, the small pieces of text a model reads and writes, and requests, the calls made to a model. It does not count credits or money. This page helps you read the figures, find who and which models used the most, and export the report.

Before you start

Seeing usage is a permission of your role. Members hold it by default; if your role does not carry it, the page says so.
Open Governance › Operations, then the Usage tab.
Nothing can be changed here: it is a page for reading.

Read the workspace's usage

Open the Usage tab.
Choose the range: Today, 7 days or 30 days.
Read the four summary tiles:
Total tokens, with the exact figure.
Requests.
Avg tokens / request.
Cache hit rate
the share of prompt tokens served from cache.

The four summary tiles: Total tokens, Requests, Avg tokens / request and Cache hit rate. — 01The tiles sum up the range you picked.

Switch between By agent and By model.
The list shows each row's requests and a bar from prompt, what the model read, to completion, what it wrote, with the cached part of the prompt.

Usage over 7 days, By agent, for Harbor & Pine: Silvestro, Pixel, Luna, Lumen and Platform services. — 02By agent shows which agents made the most model calls.

In By agent, model calls not run by any agent, such as Studio generations and schedulers, are gathered under Platform services.

Find the heaviest users

In By agent, look at the top of the list: those agents used the most tokens in the range.
In By model, see which models carried the load.

The same range, By model. — 03By model shows which models served the work.

For an explanation of spikes and outliers, use the Ask Chobin footer at the bottom of the tab. See Asking Chobin.

Export the report

Click Export as document.
Keep the title, which starts as LLM usage report · and the range, or change it.
Choose PDF, DOCX, XLSX, HTML, MD or TXT, and click Save.

The Save as document dialog for "LLM usage report · Last 7 days". — 04The report holds the summary and the full breakdown.

The report holds the summary and the breakdown table by agent or by model: Requests, Prompt, Completion, Total, Cached and Cache hit. It is saved as a document in your private Library space.

What you will see

No usage in this range

no data on a tile without data.

You don’t have the permission to see usage — Your account does not hold the permission to see this workspace’s usage. An administrator can grant it.

The refusal card for an account without the permission to see usage. — 05Without the permission, usage stays closed.

Usage analytics isn't available.

Could not load usage analytics.

Good to know

Tokens are not credits. Each model spends credits at its own rate. The credits used and left for the period are in Settings › Billing.
Work on your own key or through Nexa Bridge also appears here as tokens. It never spends credits, but counts in the daily allowance for your own models.
The cache hit rate is reported only for models whose provider reports it.
For one specific run, open its receipt.