At AgentCon Boston, I delivered a packed 2-hour hands-on workshop—so full we needed overflow seating—focused on building real, production-ready AI agents using Azure AI Foundry, Semantic Kernel, and LangChain.
To ensure everyone could follow along without setup issues, attendees were provided pre-configured Windows VMs, each loaded with the required SDKs, repo, tools, and environment variables. This made the entire experience smooth, consistent, and truly hands-on from the very first step.
We began inside Azure AI Foundry, where participants built an AI agent using prompt orchestration, tools, memory, and multi-turn planning. Within minutes, everyone had a functioning agent running in the portal capable of planning tasks, executing tools, and managing multi-step interactions without custom glue code.
Next, we brought that agent into real applications. I demonstrated how to call the same agent programmatically using Semantic Kernel and LangChain, showcasing how both frameworks support agentic workflows—multi-step reasoning, tool execution, and structured retrieval. Developers could immediately see how to fit these frameworks into their existing stacks.
We then dove into agentic RAG and enterprise retrieval, integrating Azure AI Search and vector embeddings to enable grounded, domain-aware responses. We covered how to apply secure enterprise access patterns so agents only retrieve content users are permitted to see—critical for compliance and real-world deployments.
The workshop concluded with a live end-to-end demo of a fully integrated agent retrieving information, reasoning over it, and taking action. Everyone left with a ready-to-clone repository and a complete blueprint for building actionable agents.
Workshop materials are available here: https://github.com/nhcloud/agentcon-workshop
A massive thank-you to everyone who attended—especially those standing in the back once the room filled up. Your energy made this session a standout moment of AgentCon Boston.