Token analysis

Token analysis aggregates calls over a window by one trace or business dimensions so you can see "what one user or one agent did in sequence".

Entry

After sign-in, open Observability → Token analysis at /trace-clues.

How to get aggregatable data

The platform relies on dimension headers on requests (e.g. x-trace-id, business user, app name). When integrating:

  • Use the same x-trace-id (or your equivalent) for one user session or one business flow;
  • Put stable, searchable values in x-user-id, x-agent-name, etc. (same as defaultHeaders in SDK quickstart).

Then by trace / by user / by agent views and drill-downs stay meaningful.

Common actions in the UI

  • Time range: presets or custom interval, similar to log pages.
  • Aggregate dimension: by trace, user, agent, etc. (per current console).
  • Drill-down: from the list into a timeline / detail page (often /trace-clues/detail with query params).

See also

Back to docs home