Usage Statistics
The Organization → Statistics tab gives an overview of the organization's analysis volume, independent of its cost (see AI Cost Tracking for the financial side).
Available metrics
- Jobs — total number of analyses, completed, failed, in progress.
- Tokens per user — token consumption (input/output) broken down by account.
- Tokens per pipeline — consumption broken down by pipeline (with a dedicated line for analyses not tied to a specific pipeline).
The display toggles between a "by user" view and a "by pipeline" view, shown as a chart and a table.
Portfolio calls by channel
The Portfolio calls section counts pipeline calls by channel:
| Channel | What is counted |
|---|---|
| Web interface | documents submitted and chat messages sent from the application |
| Teams | documents and messages received by the Microsoft Teams bot |
| API | calls through an API key (POST /api/jobs, conversations) |
| MCP | calls to the submit_document and ask tools of a pipeline exposed over MCP |
A call is a submitted document or a chat message; Designer trials, tests and benchmarks are not calls. The period is picked at the top (7, 30, 90 days or 12 months). For each channel, the tile shows the total and, when it differs, how many calls completed.
The chart shows the day-by-day trend, stacked by channel; hover over a column to read the day's details. The breakdown table gives the same figures by pipeline, by caller (user, or API key name) or by team (API keys form a group of their own).
In the cloud edition, API and MCP calls use call credits (see Call credits): this view helps anticipate usage and spot abnormal activity (a human account chaining hundreds of calls, for instance).
Analyses made before channels existed were classified on a best-effort basis: API and MCP are not told apart there (everything shows as API), and some old Designer chat trials may show up as web.
This page is restricted to management profiles: a regular user only sees the list of their own analyses, not aggregated statistics.