Microsoft’s Agentic AI Frameworks: AutoGen and Semantic Kernel
Table of Contents
Semantic Kernel Overview
Semantic Kernel is an enterprise-ready AI framework with AI capabilities and support for building agent production applications. It includes the Semantic Kernel Process Framework for embedding AI into workflows. Customers should choose Semantic Kernel for applications that require AI capabilities with enterprise-grade support. Semantic Kernel has an Agent and Process Framework in preview, reaching version 1.0 across .NET, Python, and Java, and ready for production use. Resources for Semantic Kernel include the Microsoft Semantic Kernel GitHub Repository, Semantic Kernel Blog, and Semantic Kernel Agent Framework on Microsoft Learn.
AutoGen Overview
AutoGen is an open-source framework designed by Microsoft Research's AI Frontiers Lab for building AI agent systems. It simplifies the creation and orchestration of event-driven, distributed agentic applications supporting multiple LLMs and SLMs, tools, and advanced multi-agent design patterns. AutoGen encompasses scenarios where multiple agents interact to complete complex tasks autonomously or with human oversight. The framework is a means for Microsoft AI Frontiers to transform cutting-edge research into agentic capabilities and facilitate the development of AI applications pushing boundaries. AutoGen encourages experimentation with innovative agentic patterns.
Collaboration Between Semantic Kernel and AutoGen Teams
The Semantic Kernel and AutoGen teams are working towards a unified multi-agent runtime, aligning AutoGen's autogen-core with Semantic Kernel for seamless transitions of agentic solutions. Customers using the multi-agent runtime from AutoGen can expect a smooth transition to Semantic Kernel in early 2025. This convergence aims to provide an enterprise-ready and support environment for customers exploring cutting-edge agentic patterns. Leveraging both Semantic Kernel and AutoGen capabilities allows developers to innovate while ensuring stability in production environments. Stay tuned for updates on the pipeline bridging AutoGen agentic applications into Semantic Kernel.