Appearance
Usage
The Usage page shows how much AI work Overdesk has done for your workspace: how many requests it made to the model on your Anthropic API key, how much text went each way, and what that cost. It is how a support lead keeps an eye on the model spend behind the drafts.
Open Usage in the sidebar. The period selector in the top right sets the window the figures cover, 7, 30 or 90 days, and requests and cost are each compared with the equally long period before. The same selector sits on Home and on Analytics, and your choice is remembered.
Beneath the figures, a line gives the all-time totals for the workspace: what has been spent since it started, across how many requests.
What the page shows
| Figure | What it counts |
|---|---|
| AI requests | Every request Overdesk made to the model in the period |
| Cost | What those requests cost at Anthropic's published prices |
| Input tokens | Text sent to the AI |
| Output tokens | Text the AI wrote back |
The count covers everything that runs on your key: reply drafts, follow-up nudges, triage, reports, analyses, Ask Overdesk answers in Slack, issue details drafted from a conversation, and investigation briefs written for a finding's issue. Indexing your knowledge base does not run on your key, so it never appears here.
The Current plan panel on the Billing page shows the same four figures for the whole lifetime of the workspace, rather than for a period.
Cost is an estimate
Overdesk works out the cost from the token counts and Anthropic's published prices. Anthropic bills you directly, so the invoice in your Anthropic account is the final figure. The page covers AI usage only: your Overdesk subscription is billed separately, through Stripe. See Managing your subscription.
Cost per draft
To see what individual drafts cost, open the Agent drafts page. Every draft shows its Model and its Cost. See Drafting model.
Analytics puts the same spend beside what it produced: how many drafts your team used, and what each one cost.
Watching spend
With Ask Overdesk, your team can ask about AI spend from Slack: see Ask Overdesk.