Q01 - Question
A team connects an agent in Foundry Agent Service to a remote MCP server by using an MCP tool object. The team must limit the agent to a subset of the server's tools and must keep a person in the loop before any tool call runs. Which two configuration options of the MCP tool connection meet these requirements? Each correct answer presents part of the solution.
Domain: Implement generative AI and agentic solutions (30–35%) Type: Multiple choice
- A. Open a manual MCP client session for every tool call made by the agent.
- B. Restrict the connection to a list of allowed tools.
- C. Rewrite the tool definitions that are hosted on the remote MCP server.
- D. Require human approval for tool calls.
B and D are correct.
Explanation: When Foundry Agent Service connects to a remote MCP server through an MCP tool object, the connection can restrict which tools are allowed and can require human approval before tool calls run.
A is incorrect: Using the MCP tool object avoids manual MCP client sessions and function-tool wrappers, so adding manual sessions works against that approach.
C is incorrect: Tool definitions remain on the remote MCP server. The connection settings control access and approval rather than changing the hosted definitions.
Q02 - Question
A web application must run a workflow that was built in Microsoft Foundry and show users the progress of completed actions, including any pause for human input, before the final response arrives. What should the developers do?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. Rebuild the workflow logic in application code and call each agent directly
- B. Copy the workflow YAML into the application and run it locally
- C. Invoke the saved workflow by name with the Azure AI Projects SDK and read the streaming events
- D. Poll the visual designer canvas for node status updates
C is correct.
Explanation: A workflow designed in Foundry can be invoked programmatically by referencing its name through the Azure AI Projects SDK, and streaming events expose progress, completed actions, pauses for human input, and the final response.
A is incorrect: Rebuilding the logic in code duplicates the workflow definition and loses the orchestration and state handling the platform already manages.
B is incorrect: The YAML view is for editing and source control of the definition, not for running the workflow inside the application.
D is incorrect: The visual designer is an authoring surface and is not a programmatic source of run progress for an application.
Q03 - Question
An architect is documenting which SDK a team should use for each part of a Microsoft Foundry chat solution. Which two statements correctly describe the SDK options? Each correct answer presents a complete solution.
Domain: Implement generative AI and agentic solutions (30–35%) Type: Multiple choice
- A. The OpenAI SDK provides access to project-level features such as agents, evaluations, tracing, and connections.
- B. The Foundry SDK provides access to project-level features such as agents, evaluations, tracing, and connections.
- C. Using the Foundry SDK removes the need to deploy a generative model in the Microsoft Foundry project.
- D. The OpenAI SDK provides model inference with full OpenAI API compatibility.
B and D are correct.
Explanation: The Foundry SDK gives you project-level features such as agents, evaluations, tracing, and connections, while the OpenAI SDK provides model inference with full OpenAI API compatibility. Choose the SDK based on which of these capabilities the application needs.
A is incorrect: Project-level features such as agents, evaluations, tracing, and connections come from the Foundry SDK, not the OpenAI SDK.
C is incorrect: You still deploy a generative model in the Microsoft Foundry project before an application can chat with it.
Q04 - Question
A finance team compiles monthly reports manually. A staff member reads incoming data, decides which items belong in the report, and completes the follow-up steps. The team wants to replace this manual process with an AI agent built on Microsoft Agent Framework. Which statement correctly describes the role the AI agent plays in this solution?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. The agent stores the incoming data so a staff member can review and act on every record.
- B. The agent runs fixed rule-based scripts to process the data without using a generative AI model.
- C. The agent uses generative AI to interpret the data, make decisions, and complete the tasks with minimal human intervention.
- D. The agent deploys a model endpoint that other applications call to generate the report.
C is correct.
Explanation: AI agents use generative AI to interpret data, make decisions, and complete tasks with minimal human intervention.
A is incorrect: Storing data for manual review does not minimize human intervention.
B is incorrect: AI agents rely on generative AI models, not just fixed rule-based scripts.
D is incorrect: Deploying a model endpoint is a hosting task, not the role of the AI agent interpreting the data.
Q05 - Question
You are configuring an AI agent to call an internal pricing API and search files hosted by a model provider. You use the Microsoft Agent Framework to minimize custom code. Which two features of the framework support these requirements? Each correct answer presents part of the solution.
Domain: Implement generative AI and agentic solutions (30–35%) Type: Multiple choice
- A. The framework restricts tools to provider-hosted capabilities.
- B. The framework generates the schema for custom Python function tools and orchestrates their invocation.
- C. The framework requires developers to parse the model output and invoke functions manually.
- D. The framework supports provider-hosted tools, such as file search, alongside custom function tools.
B and D are correct.
Explanation: Tools let an agent take actions such as calling APIs or searching files. The framework supports provider-hosted tools and custom Python function tools, and it generates the schema and orchestrates invocation automatically.
A is incorrect: The framework supports custom Python function tools, so it is not restricted to provider-hosted capabilities.
C is incorrect: The framework orchestrates tool invocation automatically, so developers do not need to parse output and invoke functions manually.
Q06 - Question
You are designing a solution in which several agents discuss a proposal together, a human reviewer can be asked for input during the discussion, and the conversation must stop when the topic is resolved. Which component of group chat orchestration provides this control?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. The first agent added to the conversation
- B. An edge defined between two executors
- C. The chat client used by each agent
- D. The chat manager
D is correct.
Explanation: In group chat orchestration, a central chat manager controls the flow. It controls turns, can request human input, and determines when the discussion ends.
A is incorrect: Turn order is decided by the chat manager, not by the order in which agents were added.
B is incorrect: Edges connect parts of a workflow, but they do not manage turn taking or end a group conversation.
C is incorrect: A chat client connects an agent to an AI model; it does not coordinate the conversation among multiple agents or request human input.
Q07 - Question
Your organization runs several agents that were built by different teams on different platforms. The teams want a common way for those agents to find each other, exchange information, and run tasks together instead of writing custom integration code for every pair of agents. Which approach addresses this requirement?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. Merge all agent logic into a single Azure AI Agent so that no cross-agent communication is required.
- B. Store each agent's configuration in a shared file location that every team reads at startup.
- C. Adopt the Agent-to-Agent (A2A) protocol as a standardized framework for agent discovery, communication, and coordinated task execution.
- D. Replace the agents with scheduled batch jobs that write results to a common database.
C is correct.
Explanation: The evidence supports C because the A2A protocol is a standardized framework for agent discovery, communication, and coordinated task execution.
A is incorrect: Merging all logic into a single agent removes the multi-agent design and does not provide a standard way to work with agents on other platforms.
B is incorrect: Storing configuration in a shared file does not define how agents discover capabilities or exchange requests and results at runtime.
D is incorrect: Replacing agents with batch jobs does not provide agent discovery or coordinated task execution between agents.
Q08 - Question
You are planning a production chat application that calls a model deployed in a Microsoft Foundry project. You must decide how the application authenticates to the endpoint. Which approach should you plan for?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. Embed the endpoint URL in the client application and call the endpoint without credentials.
- B. Use Microsoft Entra ID authentication, which is generally used for production applications.
- C. Use the Model playground session to authenticate the production application at runtime.
- D. Authenticate through the project endpoint only, because the Azure OpenAI endpoint does not accept application calls.
B is correct.
Explanation: A Foundry project exposes both a project endpoint and an Azure OpenAI endpoint, and production applications generally use Microsoft Entra ID authentication to call them.
A is incorrect: Calls to a project endpoint or an Azure OpenAI endpoint require credentials, so an unauthenticated client is not a valid design.
C is incorrect: The Model playground is for interactive testing before you write code, not for authenticating a running application.
D is incorrect: Each project has both a project endpoint and an Azure OpenAI endpoint, and applications can be built against either one.
Q09 - Question
You are starting a new Microsoft Foundry chat application that supports multi-turn conversations. You want the service to track conversation state instead of building that logic in your application code. Which API should you use?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. The ChatCompletions API, because it stores conversation state in the project endpoint.
- B. The Model playground code sample, called directly as the production API.
- C. The azure-ai-projects package alone, without a chat API.
- D. The Responses API, because it provides stateful multi-turn response generation and is the recommended approach for new applications.
D is correct.
Explanation: The Responses API provides stateful multi-turn response generation through an OpenAI-compatible client, and it is the recommended approach for generating AI responses in new Microsoft Foundry applications.
A is incorrect: The ChatCompletions API does not provide stateful response tracking, so your application must manage the conversation history.
B is incorrect: Playground code samples are a starting point for development and are not a separate API.
C is incorrect: Project-level SDK access does not by itself generate chat responses, so you still choose a chat API.
Q10 - Question
A company runs several Azure AI Agents that use the same set of tools. Tools are added, updated, and removed frequently, and each change currently requires editing hardcoded integrations in every agent. Which approach reduces this maintenance work?
Domain: Implement generative AI and agentic solutions (30–35%) Type: Single choice
- A. Use dynamic tool discovery so each agent queries the MCP server for the current tools at runtime.
- B. Copy the tool definitions into each agent project and update the copies on a fixed schedule.
- C. Reduce the number of tools so that the hardcoded integrations change less often.
- D. Store the tool definitions in the agent instructions so the model can read them during a run.
A is correct.
Explanation: Dynamic tool discovery lets an agent query an MCP server for the current tools instead of hardcoding integrations. Tool updates are centralized on the server, and the agents adapt as tools are added, updated, or removed.
B is incorrect: Duplicated definitions still need to be edited in every agent, which is the maintenance problem you are trying to remove.
C is incorrect: Limiting the tool catalog does not address the hardcoded integrations that must be changed whenever a tool changes.
D is incorrect: Placing definitions in instructions keeps the tool list tied to each agent instead of centralizing it on the MCP server.