S06 Patrones Multiagentes

In a connected-agent architecture, which responsibilities belong to the main agent?

• Execute domain tasks; return results; use tools if needed

• Interpret intent; choose connected agent; pass context; aggregate results

• Provision identity and secrets; rotate keys; manage quotas

• Host UI; collect telemetry; render charts

You need to build an enterprise-grade AI agent on Azure with broad model choice, data grounding, and enterprise security (Entra, governance). Which option best fits?

• OpenAI Assistants API

• AutoGen

• Azure AI Foundry Agent Service

• Microsoft 365 Agents SDK

A business team with no coding experience wants to create simple agents that automate daily tasks inside Microsoft 365. What should they use?

• Copilot Studio agent builder

• Semantic Kernel Agent Framework

• Microsoft 365 Agents SDK

• AutoGen

What is the correct high-level sequence to stand up a connected multi-agent solution?

• Create main agent → Run → Create thread → Add tools → Create connected agent

• Initialize client → Create connected agent → Register as ConnectedAgentTool → Create main agent with tools → Create thread & message → Run → Handle results

• Create thread & message → Run → Create agents → Attach tools → Initialize client

• Register tools → Create thread → Initialize client → Create agents → Run

Which is true about AutoGen?

• Supports only Java and Go; no Azure integration

• Node.js only; not suited for long-running agents

• Python only; can’t use multiple models

• Supports C# and Python; integrates with Azure AI Foundry for enterprise security/governance (e.g., Entra Agent ID)