Building AI agents is exciting — until you hit the part where you have to wire up orchestration, manage multi-turn state, handle tool calls, containerize everything, and figure out how to deploy it securely at scale.
This session cuts through that complexity. You'll learn how to build agentic applications using the GitHub Copilot SDK — a production-ready runtime that handles the hard parts of agent execution so you can focus on what your agent actually does. Then you'll take that agent all the way to production using Microsoft Foundry Hosted Agents, with built-in identity, per-session isolation, and observability included.
The session covers the full developer journey: from writing your first agent locally, to adding tools and multi-turn memory, to containerizing and deploying with a single command.

What You Will Learn

- How the GitHub Copilot SDK works and where it fits in the agent ecosystem
- How to build a custom agent with tools, MCP integration, and multi-turn session memory
- How to containerize your agent and deploy it to Microsoft Foundry Hosted Agents
- How to choose the right hosting model for your agentic workloads