6:00 - 6:30 PM: Open Mic - Share what you are working on or learned / lightning talks + social time 6:30 - 7:30 PM: Presentation ------------------------------------------------------------------------------- AI agents become useful when they can reach real application capabilities and collaborate across system boundaries. In .NET, that requires more than adding function calling to a chat loop.
In this code-first session, we will discuss how to build an end-to-end agentic system in C#. A .NET MCP server will expose strongly typed application tools and contextual resources. A specialist agent built with Microsoft Agent Framework will use those capabilities and be published through A2A. A coordinating .NET application will discover the remote agent, delegate work, stream its response, and retain deterministic control over workflow and approval decisions.
Along the way, we will answer the practical architecture questions: When should something remain a C# method? When should it become an MCP tool? When should it become an A2A agent? We will also examine tool contracts, agent discovery, state, authorization, failure handling, observability, and human approval. Attendees will leave with a reusable architecture for building interoperable agentic services in .NET without turning every API, service, and workflow into an agent.
