Microsoft Launches Open-Source Agent Framework for AI Agents and Workflows
Microsoft has launched the Microsoft Agent Framework, an open-source SDK and runtime environment designed to create and manage AI agents and complex multi-agent workflows. This new tool enables developers to build both simple chat agents and intricate, graph-based orchestrations.
The Agent Framework offers flexibility in deployment, allowing agents to run in containers, on-premises, or across multiple clouds. This ensures portability across different environments.
Microsoft has integrated the framework with Azure AI Foundry, Microsoft 365 Copilot, and other agent platforms, enabling seamless connectivity with multi-agent systems. It supports the Model Context Protocol (MCP) and Agent2Agent (A2A) protocol, facilitating dynamic discovery and invocation of external tools or data servers, and collaboration across different runtime environments.
Developed by Microsoft teams including Microsoft Research and Azure AI Foundry, the framework unifies and extends concepts from the Semantic Kernel and AutoGen projects. It supports both .NET and Python, and any REST API with an OpenAPI specification can be instantly imported as a callable tool. Developers can experiment locally and deploy applications to the Azure AI Foundry with features like observability, high availability, and compliance.
The Microsoft Agent Framework, an open-source SDK and runtime environment, empowers developers to create and manage AI agents and complex multi-agent workflows. Its flexibility, connectivity, and ease of use make it a valuable tool for building and deploying AI solutions.
