Krux
April 5, 2026
Microsoft's Framework Lets One Workflow Use Seven AI Models
Published: April 5, 2026 at 12:26 AM
Updated: April 5, 2026 at 12:26 AM
100-word summary
Microsoft just shipped Agent Framework 1.0, and the headline feature is simple: mix and match AI models from OpenAI, Anthropic, Google, Amazon, and others in the same workflow. A writer agent powered by Claude can pass its draft to a reviewer agent running on GPT-4, all without rewriting your code. The framework supports .NET and Python, ships with a browser-based debugger called DevUI that shows every decision your agents make in real time, and includes migration guides for anyone stuck on Semantic Kernel or AutoGen. The surprise? Microsoft is betting developers want model flexibility more than model loyalty.
What happened
Microsoft just shipped Agent Framework 1.0, and the headline feature is simple: mix and match AI models from OpenAI, Anthropic, Google, Amazon, and others in the same workflow. A writer agent powered by Claude can pass its draft to a reviewer agent running on GPT-4, all without rewriting your code. The framework supports .NET and Python, ships with a browser-based debugger called DevUI that shows every decision your agents make in real time, and includes migration guides for anyone stuck on Semantic Kernel or AutoGen. The surprise?
Why it matters
Microsoft is betting developers want model flexibility more than model loyalty.