AzureAIAgentClient vs AzureAIClient

[Javier-Jimenez99]

Hello,

I am working with the Azure AI Foundry Agent Framework to create hosted agents. I have a few questions regarding the behavior and differences between two classes: AzureAIClient and AzureAIAgentClient.

The Context:
I typically use AzureAIClient to instantiate agents on Foundry. However, I’ve noticed several official samples (like this one) use AzureAIAgentClient instead.

The Confusion:

1. Redundancy: According to some documentation, AzureAIAgentClient is supposed to automatically create a Foundry agent when it is run. This seems counter-intuitive: if I am writing code to manage the logic of a hosted agent, why would that code trigger the creation of a new agent entry in Foundry every time it runs?
2. Inconsistent Behavior: I attempted to replicate the approach used in the linked sample. While the "hosted agent" is created, running the code does not result in a new agent being created in the portal, which contradicts what I understood from the documentation.

My Questions:

• What is the specific architectural difference between AzureAIClient and AzureAIAgentClient?
• Is AzureAIAgentClient actually intended to provision a new agent resource dynamically, or is it simply a wrapper for interacting with an existing one?
• Is it possible the documentation is outdated, or am I potentially missing a configuration flag that triggers the agent creation?

Any clarification on the intended "Developer Flow" for these two classes would be greatly appreciated.