Microsoft’s Agent Framework reaches version 1.0

Software development has been transformed by AI and especially AI agents, which are capable of (semi)autonomous work in all areas of software development.

About two weeks ago, Microsoft made available the first "RTM" version of Microsoft Agent Framework ("MAF"), which was announced in October. I wrote about this in my blog earlier.

The Agent Framework makes it very easy to use languages like C# and Python to use different agents easily. Similarly, creating your own agents becomes almost trivial: you tell the framework what kind of agent you want, which is the base model (LLM), and what kind of "personality" you want to give it.

Many features are supported, including:

• Service Connectors
• Middleware Hooks
• Agent Memory
• Multi-Agent Orchestration
• Declarative Agents and Workflows (as YAML script)
• A2A (Agent-to-Agent) and MCP (Model Context Protocol) support
• Migration assistants for people moving from Semantic Kernel and AutoGen.

Very exciting!