Skip to content

AI Studio

AI Studio runs are rule executions that use a model and spend credits. Two read commands expose that usage, so you can report on credit spend or poll for new runs without opening the Asana admin console.

Terminal window
cyber-asana ai-studio runs
cyber-asana ai-studio runs --workspace-gid <gid>
cyber-asana ai-studio runs --start-at 2026-01-01T00:00:00Z --end-at 2026-02-01T00:00:00Z --all
Command Arguments Options
runs — --workspace-gid <gid> / --workspace <gid>, --division-gid <gid>, --start-at <datetime>, --end-at <datetime>, pagination

--workspace-gid falls back to ASANA_WORKSPACE_GID. --division-gid scopes the read to one division; without it Asana uses the organization’s first licensed division.

Each row names the rule, who triggered it, the model, the status, and the credits used. Runs come back oldest first.

--start-at is inclusive and --end-at is exclusive. Both are ISO 8601 date-times and filter on when the usage was recorded.

The endpoint returns a fixed row shape, so runs takes no --opt-fields.

Because runs are oldest first, the last row’s timestamp is a cursor. Pass it as --start-at on the next call to read only what is new:

Terminal window
cyber-asana ai-studio runs --start-at 2026-01-15T09:30:00Z --all --toon
Terminal window
cyber-asana ai-studio seats
cyber-asana ai-studio seats --state active
cyber-asana ai-studio seats --state revoked --json
Command Arguments Options
seats — --workspace-gid <gid> / --workspace <gid>, --division-gid <gid>, --state <state>, pagination

--state is active or revoked. Each row shows the user, the license tier, the state, and when the seat was assigned. The list is a current snapshot, not a history.