Building an agent that answers questions on a database with two tables is easy. But what if you have hundreds? In this practical workshop, we will move beyond the "Hello World" of AI to build a robust SQL Agent using the Microsoft Agent Framework. You won't just look at slides; you will write the C# code to solve the engineering challenges that arise when scaling from a fragile demo to a real-world application. We will start with a basic agent and iteratively improve it to handle complexity, focusing on: Scalable Tool Design: Building granular functions that provide high-signal context without confusing the model. Advanced Context Management: Implementing Just-in-Time (JIT) retrieval and structured note-taking to handle long-running tasks without blowing up token limits. Operational Guardrails: Configuring system instructions to keep your agent focused and predictable. Prerequisites: - .NET 10 installation - Docker installation (other container runtime is fine e.g. Podman) - Permissions to install NuGet packages (Azure OpenAI SDK, Microsoft Agent Framework, Microsoft Extensions AI, Microsoft Extensions AI Evaluation, Dapper) Azure accounts will be provided to you.