Monitor agent flows
Unlock the complete study guide + 1,040 practice questions across 16 full exams.
Bundled into the existing Designing and Building Integrated AI Agent Solutions in Copilot Studio premium course — no separate purchase.
14-day money-back guarantee — no questions asked.
Included in this chapter:
- Two objects, two altitudes: the monitoring grid
- The agent flow's three tabs, and what each one owns
- Working an agent flow's run history
- Before any run: the authoring-time surfaces
- Real-time activity map or historical Activity page
- Reading an activity: nodes, statuses, and Rationale
- Visibility: who sees which activity and which analytics
- Correlating an agent activity with a flow run
- How these choices show up in questions
Which monitoring surface answers which question
| Surface | Object it reports on | Altitude | Evidence you get | Wrong choice when |
|---|---|---|---|---|
| Flow checker (designer) | Agent flow definition | Outside the grid: no execution | Every error and warning in the flow as authored; errors block publishing | The flow already published and fails only for some inputs |
| Agent flow Overview tab | Agent flow | Outside the grid: configuration | Name, description, connections, owner and co-owners, created and modified dates, most recent runs, delete or turn off | You need the error inside one of those runs |
| Agent flow Activity tab | Agent flow | One execution | Run history with status and duration; select a run to step through each action's inputs and outputs; resubmit or cancel selected runs | The question is which tool the orchestrator picked, or why |
| Agent flow Analytics tab | Agent flow | Aggregate | Number of runs, failure rate and performance trends; solution flows only, up to 30 minutes behind | You need one run's action, payload or error |
| Real-time activity map (Test your agent) | Agent | One execution, reproduced now | The generated plan, the selected action, parameter values, outputs, per-step duration, highlighted errors | The failure already happened and cannot be reproduced |
| Agent Activity page (historical) | Agent | One execution, already past | Recorded activities with status, completed steps and last step; filter by Failed; transcript and map views; Rationale on demand | You need step-level detail inside the flow the agent called |
| Agent Analytics page | Agent | Aggregate | Engagement and adoption, conversation or run outcomes, tool use and success rate, knowledge source use, satisfaction, over the reporting period | You are diagnosing a single conversation or a single flow run |
Decision tree
Cheat sheet
Unlock with Premium — includes all practice exams and the complete study guide.
Also tested in
References
- Agent flows overview - Microsoft Copilot Studio
- Review agent activity - Microsoft Copilot Studio
- Monitor your agent flows - Microsoft Copilot Studio
- Agent flows in Microsoft Copilot Studio FAQ FAQ
- Manage and monitor agent flows - Microsoft Learn training
- Edit and manage your agent flow in the designer - Microsoft Copilot Studio
- Analyze conversational agents - Microsoft Copilot Studio
- Analyze autonomous agents - Microsoft Copilot Studio
- Share agents with other users - Microsoft Copilot Studio
- Analytics overview - Microsoft Copilot Studio
- FlowActionBadRequest error fix for agents - Copilot Studio troubleshooting