Microsoft Designing and Implementing Multi-Agent AI Solutions : AI-500

You have a Microsoft Foundry project that contains an incident triage agent.
You have a Model Context Protocol (MCP) server registered in the organizational tool catalog. The MCP server exposes two tools named docs_search and deployment_delete.
You need to ensure that the agent can only invoke docs_search.
What should you configure?

• A. the agent instructions
• B. the agent tool configuration
• C. the agent run configuration
• D. the project details

You have a Microsoft Foundry multi-agent solution that includes a orchestrator and agents named Agent1 and Agent2. The solution has the following characteristics:
* The orchestrator routes tasks to both agents.
* The agents invoice Model Context Protocol (MCP)-hosted tools.
* Agent? sends model-generated JSON parameters directly to tool endpoints.
* The agents use a shared managed identity that has contributor access to the resource group that contains the tool-backed Azure resources.
You need to implement a guardrail strategy for tool calls. The solution must meet the following requirements:
* Prevent invalid or out-of-range tool parameters from reaching the tool endpoints.
* Scope each agent ' s permissions to its assigned tools
* Limit Agent? to refund-processing tools only.
* Log tool invocations for audit purposes.
What should you do?

• A. Configure per-agent Agent-to-Agent (A2A) capability cards, validate conversation history during orchestration, and assign the shared managed identity that has scoped Azure role-based access control (Azure RBAC) roles.
• B. Configure per-agent tool access policies, validate typed payload schemas at the tool boundary before execution, assign per-agent managed identities that have scoped roles, and enable Azure Monitor logging for tool invocations.
• C. Configure all the tools as built-in function tools, validate tool outputs after execution, and assign per- agent managed identities that have contributor access to the resource group.
• D. Configure per-agent system instructions, validate model responses after tool execution, and assign the shared managed identity that has scoped custom deny assignments.

You have a LangGraph workflow in Microsoft Foundry that is compiled as app by using a checkpointer. Each request includes a value named ticket_id.
You need to instrument the workflow so that each streamed run sends OpenTelemetry traces to Observability in Foundry. The solution must meet the following requirements:
* Correlate graph steps and tool calls for each request by the supplied ticket__id.
* Use the Azure Al OpenTelemetry tracer with the LangGraph invocation.
How should you complete the code? To answer, select the appropriate options in the answer area NOTE: Each correct selection is worth one point.

You have a multi-agent solution hosted in an Azure App Service plan.
You use Azure App Configuration to store runtime settings.
A release adds a new planner agent and changes the prompt routing
You need to recommend a release plan that meets the following requirements:
* Slowly rolls out the release to users
* Uses end-user health alerts as rollout gates
* Increases exposure without redeploying the application
* Limits production exposure until the health metrics remain green
* Implements the solution without deploying a second environment
What should you recommend? To answer, select the appropriate options in the answer area. NOTE: Each correct selection is worth one point.

You need to recommend a Microsoft Foundry multi-agent solution that has the following domain-specific requirements:
* Agent responses must never include the names of two specific sanctioned companies.
* Agent responses must NOT expose customer account numbers
Which guardrail strategy should you include in the recommendation?

• A. Create a blocklist that applies to agent outputs. Configure Personally Identifiable Information (Pll) detection at the output intervention point.
• B. Create a blocklist that applies to user input. Configure Personally Identifiable Information (Pll) detection at the user input intervention point.
• C. Apply the built-in blocklist to user inputs. Configure Task Adherence at the tool call intervention point
• D. Apply the built-in blocklist to agent outputs. Configure Task Adherence at the tool call intervention point.