Mark session as read
Record that the calling user has read the session as of now, clearing its unread indicator. Idempotent: repeated calls only move the read receipt forward.
Authentication
Path parameters
Headers
Response
Timestamp when the session was created (ISO 8601)
Unique identifier of the session asset (e.g., ‘asset_abc123’)
Canonical user-facing session status: idle, active, needs_input, or error
Timestamp when the session was last updated (ISO 8601)
Source AOP asset identifier for an AOP/workflow run
Asset ID of the collab agent the session was created with, when one was bound; null for stock-agent sessions
Canonical failure reason for failed runs; null for non-failed runs
Whether this is a branched sub-session of another session
Whether the session is unread for the calling user: it changed since they last opened it, or they never opened it and it did not originate from the web app
Plain-text preview of the most recent message, when available
Asset ID of the parent session for sub-sessions
Canonical latest-run status: scheduled, queued, running, needs_input, completed, failed, or canceled
Kind of session: ‘session’ (chat), ‘video_session’, ‘desktop_session’, or ‘mobile_session’
Channel the session originated from (e.g., ‘web’, ‘api’, ‘agent_email’, ‘agent_slack’, ‘agent_sms’)
Total LLM cost of the session in USD, when tracked. Null unless the caller is Athena staff or an owner or manager of the session’s workspace. Already includes the workspace’s LLM cost multiplier, applied once to the provider’s rate; never apply it again. Sessions priced before the in-house token ledger carry their legacy figure, without it. See docs/SESSION_TOKEN_USAGE_AND_TOKEN_COST_GUIDE.md#which-cost-field-do-i-sum.
Trigger that started an AOP/workflow run, such as schedule or api
Deprecated legacy execution state. Use session_status_v2, run_status_v2, and failure_reason_v2 for status UI.
