Microsoft has launched the Agent Framework, an open-source SDK that merges the enterprise features of Semantic Kernel with the multi-agent orchestration capabilities of AutoGen, creating a unified platform for developing AI agents. The framework supports both .NET and Python, allowing developers to build AI agents and multi-agent workflows with minimal code.
Agent Framework introduces abstractions that simplify the creation of single agents or multi-agent workflows, integrating seamlessly with familiar .NET hosting patterns and offering built-in observability through OpenTelemetry. It supports various orchestration patterns—including sequential, concurrent, handoff, and group chat—and replaces AutoGen’s approach with more deterministic and controllable orchestration, providing workflows that are easier to debug and manage.
The framework’s core features include compliance with open standards such as Model Context Protocol (MCP) and the Agent-to-Agent (A2A) protocol, alongside advanced orchestration patterns. It offers extensible connectors for integrating with Azure services, Microsoft Graph, MongoDB, Amazon Bedrock, and other enterprise systems, and includes pluggable memory stores for flexibility in building sophisticated AI systems.
Production readiness is a priority for Microsoft Agent Framework, as it comes equipped with features like observability, security, and enterprise-grade capabilities like thread-based state management and type safety. Developers can deploy their agents using standard <a href="http://ASP.NET" target="_blank" rel="noopener nofollow">ASP.NET</a> patterns, fully equipped with comprehensive monitoring capabilities.
Furthermore, the framework provides a cloud-agnostic runtime with an OpenAPI-first design, facilitating local development and seamless deployment to Azure AI Foundry. As a result, major enterprises such as KPMG, BMW, and Commerzbank are already piloting the framework for various applications like audit automation, telemetry analysis, and customer support.
Microsoft has also released migration guides for developers currently using Semantic Kernel and AutoGen, ensuring a smooth transition to the new framework which focuses on providing a unified and robust environment for AI agent development.

Collections

Microsoft released Agent Framework, an open-source SDK combining Semantic Kernel's enterprise features with AutoGen's multi-agent orchestration. The framework supports .NET and Python, offering deterministic orchestration patterns, compliance with MCP and A2A protocols, and extensible connectors for Azure, Microsoft Graph, MongoDB, and Amazon Bedrock. It includes production-ready features like OpenTelemetry observability, thread-based state management, and cloud-agnostic deployment. Companies like KPMG, BMW, and Commerzbank are already piloting it for audit automation, telemetry analysis, and customer support.

Exploring Microsoft Agent Framework: A Comprehensive Tool for Developing AI Applications