Sessions
Record how AI assets are actually used, by whom, and what it costs.
Finding AI assets tells you what is installed. Sessions tell you how those assets are used. A session is one conversation between a person and an AI asset: a chat in ChatGPT, or a task given to Claude Code.
The coding agents Claude Code and Codex are recorded from the moment an endpoint enrols. AI websites in a browser are recorded once you push the browser extension. You turn any of them off, one AI tool at a time, under AI Tools. Cursor is not recorded on the endpoint. Its cloud agents are recorded through Cloud Agents (Cursor Cloud Agents).
What is recorded
For each session: the source asset, the AI provider and model, the person and endpoint, every message with its text, token counts (the small chunks of text AI providers charge by), the tool calls the AI made, and an estimated cost.
The message text is kept one step away from everything else. Reading it takes an extra confirmation and is written to the Audit Log. See Session Trace.
Where it is stored
Sessions stay in SuperAlign storage for 30 days. If you connect your own Amazon S3 or Splunk, the data lives in your system under your own retention rules.
Who can see sessions
Admins and Auditors see all sessions. Members see only sessions from their own endpoints.