Copilot Studio Agents: Tools, Monitoring & Development
12 free practice questions with explanations
12 free questions · instant explanations · no sign-up
PassNova has 12 free Microsoft PL-900 (Power Platform Fundamentals) practice questions on Copilot Studio Agents: Tools, Monitoring & Development, each with a clear explanation. Practise them in the browser with instant feedback — 100% free, no sign-up, on any device. Updated for 2026.
Copilot Studio Agents: Tools, Monitoring & Development: example questions & answers
12 worked examples with answers and explanations below. Practise them in the browser with instant feedback on every answer.
Which Copilot Studio tool category gives an agent pre-built integrations with more than 1,400 external services such as Salesforce, SAP and Microsoft SQL Server?
- AChild agents, specialized agents invoked as tools
- BCustom prompts, templates that shape an agent's output format
- CAgent flows, deterministic automations built directly inside Copilot Studio
- DPower Platform connectors✓
Answer: Power Platform connectors are pre-built integrations with over 1,400 external services, including Salesforce, SAP, ServiceNow and Microsoft SQL Server, that let an agent read from and write to those systems in real time. Agent flows are deterministic workflows built inside Copilot Studio rather than external integrations, custom prompts structure an agent's output format rather than connecting to a system, and child agents are other specialized agents invoked as tools, not connector integrations.
A maker wants every response about competing products to follow the same structured comparison format. Which Copilot Studio tool category is designed for this?
- APower Platform connectors
- BComputer Using Agents
- CMCP servers
- DCustom prompts✓
Answer: Custom prompts are maker-defined prompt templates that structure an agent's output format for a specific use case, such as always generating a structured comparison when the user asks about competing products. MCP servers expose live operations from enterprise systems rather than formatting output, Power Platform connectors integrate with external services rather than shaping response structure, and Computer Using Agents interact with legacy graphical interfaces, which is unrelated to output formatting.
An agent needs to interact with a legacy system that has no API, by simulating mouse clicks and keyboard input on its graphical interface. Which tool category provides this capability?
- AMCP servers
- BCustom prompts that only shape an agent's output format
- CComputer Using Agents✓
- DAgent flows, deterministic automations that call APIs and connectors in a fixed sequence
Answer: Computer Using Agents, or CUA, is an advanced tool category that lets an agent interact with the graphical interface of legacy systems lacking APIs by simulating mouse clicks and keyboard input. Agent flows automate steps through defined connectors and logic rather than simulated clicks, MCP servers expose structured operations from systems that already support the protocol, and custom prompts only shape output formatting, none of which reach a system through its screen.
A coordinating agent routes a user's request to a specialized agent focused on a specific domain, which then handles the request as one of the coordinator's tools. What is this specialized agent called?
- AA child agent✓
- BA custom knowledge connector
- CAn agent flow
- DA knowledge source
Answer: A child agent is a specialized agent that a primary, coordinating agent invokes as one of its tools, enabling multi-agent architectures that route requests to domain-specific specialists. An agent flow is a deterministic workflow rather than another agent, a knowledge source is content an agent searches rather than a specialist agent, and a custom knowledge connector links to an external knowledge system rather than routing requests to another agent.
What does Microsoft's first-party Dataverse MCP server let a Copilot Studio agent do at runtime?
- AReplace the need for Power Platform DLP policies on Dataverse connectors
- BAutomatically publish the agent to every configured channel at once, without any maker review or approval
- CTrain a new foundation AI model using the organization's Dataverse tables
- DDynamically call operations such as listing tables, reading data, and creating or updating records✓
Answer: The Dataverse MCP server exposes nine operations, including listing tables, describing a table's schema, reading data, and creating or updating records, that the agent's AI orchestrator can call dynamically based on what the user asks. It does not publish an agent to channels, train foundation models on Dataverse data, or remove the need for DLP governance, since MCP-connected agents remain subject to the same DLP policies as any other connector-based tool.
Why are Copilot Studio agents that use MCP servers still subject to Power Platform DLP policies?
- ABecause DLP policies apply only to knowledge sources, and MCP servers are classified as knowledge sources
- BBecause Microsoft manually reviews each MCP server connection before it is approved
- CBecause MCP servers are classified as a type of agent flow, and agent flows fall under DLP by default
- DBecause MCP server connections rely on Power Platform connectors for their underlying connectivity✓
Answer: MCP server connections use Power Platform connectors for their underlying connectivity, so if an administrator blocks a connector that an MCP server relies on, the agent cannot use that MCP server's tools. MCP servers are a distinct tool category from agent flows rather than a type of them, Microsoft does not manually review each connection since discovery is automatic, and DLP governs connectors used as tools as well as knowledge sources, not knowledge sources alone.
What best describes an agent flow in Copilot Studio?
- AA deterministic automation workflow that an agent invokes as a tool to execute a fixed sequence of steps✓
- BA live connection to an external MCP server that exposes discoverable operations
- CA conversation path defined within a topic that the agent's intent-recognition model follows in exactly the same order every time
- DA dashboard view that summarizes an agent's total sessions and resolution rate
Answer: An agent flow is a deterministic, rule-based automation workflow built directly in Copilot Studio that an agent can invoke as a tool, ideal for processes that must always execute in a specific order or apply fixed business rules. A topic's conversation path is authored dialogue logic rather than a callable automation, an MCP server exposes external operations rather than running Copilot Studio's own workflow steps, and an analytics dashboard reports on usage rather than performing an action.
An agent flow pauses execution to request sign-off from a designated manager in Outlook or Teams before continuing. What agent flow capability is this?
- AA trigger type that runs the flow on a fixed daily schedule
- BAn approval gate supporting a human-in-the-loop process✓
- CA Power Fx expression that calculates a conditional value
- DA variable that passes an order number into the flow
Answer: Approval gates let an agent flow pause execution to request approval from a designated user through Outlook or Teams before continuing, supporting human-in-the-loop processes such as manager sign-off. A scheduled trigger only controls when a flow starts running, a Power Fx expression performs calculation or branching rather than pausing for a person, and a variable simply passes a value into or out of the flow rather than requesting anyone's approval.
A Copilot Studio evaluation checks whether the agent's response includes an exact required policy phrase every time. Which grading method is being used?
- AText match✓
- BQuality
- CTool use
- DSimilarity
Answer: Text match checks whether specific text strings, such as a required policy reference or key phrase, appear in the agent's response, which is exactly what confirming an exact required phrase calls for. Quality instead uses an AI model to judge whether a response is accurate, relevant and well-grounded, similarity compares meaning through semantic closeness even when the wording differs, and tool use validates whether the agent called the right tool or topic rather than checking response text.
On a Copilot Studio agent's analytics Overview page, which metric specifically measures the percentage of sessions handed off to a human agent or a defined escalation path?
- AResolution rate
- BAbandonment rate
- CEngagement rate
- DEscalation rate✓
Answer: Escalation rate is the percentage of sessions handed off to a human agent or escalated through a defined escalation path, distinct from the other Overview metrics. Resolution rate instead measures sessions resolved without escalation or repeated questions, engagement rate measures sessions with at least one substantive response, and abandonment rate measures sessions the user left before their question was resolved.
A maker checks the Active Users page for an agent but sees no daily or monthly active user data at all. What is the most likely reason?
- AThe agent is not configured with Microsoft Entra ID authentication, which active user metrics require✓
- BThe agent uses topic-based orchestration instead of generative orchestration
- CThe agent has not yet been connected to any knowledge source, so it has nothing to retrieve or report on for users
- DThe agent has not been added to the Copilot Studio Kit's Agent Inventory
Answer: Active user metrics such as daily active users and monthly active users require the agent to be configured with Microsoft Entra ID authentication, so without it the Active Users page has no identified users to count. Missing knowledge sources or using topic-based rather than generative orchestration would affect how the agent answers, not whether user identities are tracked, and the Agent Inventory is a separate, tenant-wide reporting feature in the Copilot Studio Kit rather than a prerequisite for an individual agent's active-user metrics.
Microsoft's recommended agent development process moves from initial design through ongoing improvement. Which sequence of stages does it describe?
- AScope, code, compile, ship, then deprecate
- BEnvision and design, build, evaluate, deploy, then monitor and improve✓
- CDesign, followed by publish, secure, archive, and finally a full retrain of the agent's AI model
- DPlan, prototype, certify, license, then retire
Answer: Microsoft's five-stage agent development process runs from envision and design, through build and evaluate, to deploy and finally monitor and improve, turning deployment into the start of an ongoing cycle rather than an endpoint. The other sequences use stage names Copilot Studio's guidance does not use for this process, mixing in terms such as certify, license or deprecate that do not appear in the five defined stages.