About This Architecture

Azure AI Agent Runtime orchestrates intelligent agents across entry surfaces like Teams, Web APIs, and Power Apps, routing user requests through Copilot Studio and Power Automate to foundational models and custom tools. The architecture layers orchestration, knowledge retrieval via Azure AI Search with Document Intelligence ingestion, and model access through Azure AI Foundry supporting Azure OpenAI GPT, Anthropic Claude, and content safety guardrails. Data flows from user entry points through the agent framework (Foundry Agent Service or Function App) to RAG indexes, custom internal APIs, and Microsoft Graph for contextual permissions. Observability spans Application Insights and AI Foundry tracing, enabling monitoring and debugging across the entire agent lifecycle. Fork this diagram to customize agent flows, add domain-specific tools, or adapt entry surfaces for your enterprise AI deployment on Azure.

People also ask

How do I build an AI agent runtime on Azure using Copilot Studio, Power Automate, and Azure AI Foundry?

Azure AI Agent Runtime connects user entry points (Teams, Web APIs, Power Apps) through Copilot Studio and Power Automate orchestration to Azure AI Foundry models and RAG via Azure AI Search. The architecture includes Document Intelligence for ingestion, custom tools via internal APIs, Microsoft Graph for permissions, and observability through Application Insights and AI Foundry tracing.