Monitoring agent sessions and inbox traces
Written By Stanislas
Last updated 16 days ago
Overview
Sessions & Inbox provides a centralized observability hub to review conversation histories, user interactions, credit consumption, and execution traces for a specific agent.
It connects high-level chat logs with underlying technical steps, allowing you to audit user exchanges, identify reasoning patterns, and troubleshoot failed requests across all deployment channels.
Prerequisites
You must be the owner of the agent or hold administrator permissions with edit access to the agent.
The agent must be created and have received interactions across at least one channel.
Step-by-step guide
1. Access Sessions & inbox
In the left sidebar of your workspace, click Agents and select the agent you want to inspect.
In the agent configuration navigation panel, scroll down to the Observability section and click Sessions & Inbox.

The main Sessions & Inbox view displays your agent's conversation records with columns for status, channel, user, conversation title, timestamp, and AI credits consumed.

2. Search and filter sessions
In the search bar at the top left, type a username, email address, or keyword from the conversation title to filter sessions in real time.
Click the channel filter dropdown (set to All channels by default) to filter conversations by specific access points, such as Agent as App, API, Widget, Mobile, Desktop, or Trigger.

Click the date filter dropdown (set to All by default) to restrict the view to Today, This Week, This Month, or select Custom Period to choose a specific date range.

3. Filter by execution errors
Locate the Errors Only toggle switch in the top toolbar next to the refresh button.
Click the toggle to display only conversations that encountered an execution failure. The numeric badge beside the toggle indicates the total number of flagged sessions.
4. Inspect session conversation and traces
Click any row in the sessions table to open the detailed drawer on the right side of the screen.
In the Conversation tab, review the full read-only transcript of the conversation between the user and the agent, including model thinking summaries, token consumption, and the active LLM.

Switch to the Logs & Errors tab to inspect the chronological execution trace, including session initialization, user prompts, internal thinking steps, and model outputs.

If a failure occurred, check the highlighted error card and tag (such as
OUT_OF_CREDITS) to identify the exact error code and root cause.
5. Export session history to CSV
Apply any desired search keywords, channel selections, or date ranges.
Click the Export icon (download tray) located at the top right of the table next to the refresh button.
The filtered session records download immediately to your local device as a CSV file.
Practical use cases
Investigating user-reported bugs: Locate an issue reported by a team member by searching their name, open the session drawer, and review the exact prompt and response flow.
Auditing multi-channel adoption: Filter interactions by channel (such as Widget vs. API) to evaluate where your agent receives the highest engagement and how token consumption differs.
Diagnosing credit depletion: Enable the Errors Only toggle, spot interactions interrupted by
OUT_OF_CREDITS, and inspect token expenditure per turn to refine model selection or prompt length.
Tips & best practices
Use the refresh button beside the export icon to poll for new conversations without refreshing your entire browser page.
Review thinking steps in the Logs & Errors tab to ensure your agent follows complex instructions and reasoning constraints as expected.
Combine channel and date filters before exporting to produce targeted reports for team reviews or external audits.
Troubleshooting
No sessions appear in the list:
Cause: Active search terms or date ranges exclude existing conversations, or the agent has not yet received messages.
Fix: Reset the channel and date dropdowns to "All", clear the search input, and click the refresh button.
Cannot see the Observability section:
Cause: Your user account lacks permission to edit the agent or workspace admin rights.
Fix: Contact the agent owner or workspace administrator to grant edit access to the agent.
Export button does not download any data:
Cause: The active filters currently return 0 matching sessions.
Fix: Verify that at least one row is visible in the table before clicking the export icon.
Additional resources
Analytics
Lock agent
Creating an automation with agents
Was this helpful?
More in Agent Observability
Agent feedbackAgent changelogStill need help? Ask the team