Purpose of this guide: Explain four Microsoft-centered approaches to building AI agents in plain English, compare their strengths and tradeoffs, and help a reader select a sensible starting point.
Reviewed against Microsoft documentation: May 24, 2026. AI products evolve quickly, so confirm current model availability, licensing, preview status, region support, and pricing before a production decision.
Microsoft Foundry, Azure OpenAI, Microsoft Copilot Studio, and a local Python agent application are related choices, but they do not mean the same thing.
The best option is often a combination. For example, Python can contain custom agent logic while Azure OpenAI supplies a model, or custom Python agent code can eventually be managed through Microsoft Foundry.
Microsoft's newer unified Azure AI platform is documented as Microsoft Foundry. It brings agents, models, and tools into one resource and management experience. Microsoft also documents a migration or upgrade path for existing Azure OpenAI resources into a Foundry resource while preserving existing state such as endpoints and keys.
This matters because a team may see both terms:
In simple terms: Azure OpenAI describes access to OpenAI model capability in Azure; Foundry is the larger platform experience that can organize model and agent work around it.