Microsoft’s multi-agent framework, AutoGen, serves as a foundational element for various enterprise initiatives, especially following the launch of AutoGen v0.4 in January. The company is working to unify its agent frameworks and enhance observability features. Recently, Microsoft introduced the Agent Framework in public preview, which will function as the primary orchestration and agent framework moving forward.
As part of this consolidation, Microsoft announced that AutoGen and Semantic Kernel will transition to “maintenance mode.” This designation means they will not receive new features but will continue to receive essential bug fixes, security patches, and stability updates. Microsoft informed clients about the necessity of migrating to the new framework to benefit from open standards, durability, and integration with Azure AI Foundry.
The existing workloads utilizing AutoGen or Semantic Kernel will be supported without any anticipated breaking changes. This strategy indicates a focus on agentic AI that emphasizes observability and data protection within a unified platform for agent development post-deployment.
The new Agent Framework integrates capabilities from the Semantic Kernel and AutoGen into a single SDK, enabling users to develop AI agents, manage multiple agent deployments, and establish observability systems. Sarah Bird, Microsoft’s Chief Product Officer for Responsible AI, highlighted the demand for a cohesive approach that encompasses various functionalities for developers and businesses engaging with AI agents.
The framework includes five key features designed for enterprise AI agent development. These consist of local experimentation capabilities within Azure AI Foundry, API integration through OpenAPI, use of orchestration agents, reduced context switching across platforms, and the ability to construct multi-agent systems that utilize varied agent platforms.
Additionally, the framework incorporates responsible AI elements such as Task Adherence, PII Detection, and Prompt Shields to enhance safety and security. Microsoft is also committing to contributing to the OpenTelemetry standard, allowing developers using the Agent Framework to monitor attributes like quality, performance, and costs.
As the landscape of AI agents evolves, competition remains from other frameworks like LangChain, CrewAI, and LlamaIndex, all of which aim to centralize tools for building, deploying, and monitoring AI agents.
Source: https://venturebeat.com/ai/microsoft-retires-autogen-and-debuts-agent-framework-to-unify-and-govern