039 - Let's Create an Agent in Azure (Deep Dive)

This episode walks through the architecture of an agent on Microsoft Foundry, using a small IT help desk agent as the running example: a web front end, a backend that authenticates with Microsoft Entra ID, a Foundry agent running a GPT-5 deployment, and one custom tool that looks up a support ticket. Rather than a step-by-step build, it stays at the level of what each piece is and why it's there — the authentication approach that avoids API keys, the difference between a model and an agent, how the tool-calling loop works, the identities and permissions involved once an agent starts taking actions, and where a Foundry-managed prompt agent gives way to a fully hosted one. It also covers what separates a demo from something production-ready: keeping the agent's definition in source control rather than the portal, building an evaluation dataset instead of testing by hand, watching what an agent does once it's live, and treating anything a tool retrieves as untrusted data rather than a trusted instruction. Useful for anyone — technical or not — who wants a clear mental model of what "an agent" is made of, without needing to write any of the code themselves. Sources & References Quickstart: Create a prompt agent, Microsoft Foundry docs — https://learn.microsoft.com/azure/foundry/agents/quickstarts/prompt-agent Use function calling with Microsoft Foundry agents (Python), Microsoft Foundry docs — https://learn.microsoft.com/azure/foundry/agents/how-to/tools/function-calling Role-based access control for Microsoft Foundry — https://learn.microsoft.com/azure/foundry/concepts/rbac-foundry Agent evaluators, Microsoft Foundry docs — https://learn.microsoft.com/azure/foundry/concepts/evaluation-evaluators/agent-evaluators Built-in evaluators reference, Microsoft Foundry docs — https://learn.microsoft.com/azure/foundry/concepts/built-in-evaluators Voice narration is AI-generated.