Microsoft has launched the Microsoft Agent Framework to simplify the creation of advanced AI agent workflows.
The open-source tool is designed to enable developers to create complex multi-agent workflows using .NET or Python.
Microsoft has launched the Microsoft Agent Framework, an open-source SDK for creating, orchestrating, and deploying AI agents and workflows, supporting .NET and Python.
Microsoft says the framework, available on GitHub, lets developers create simple chat agents and complex multi-agent workflows with graph-based orchestration. Developers can experiment with the framework locally and then deploy their applications to Azure AI Foundry, which includes built-in observability, resilience, and compliance features. Multi-agent systems can connect to Azure AI Foundry, Microsoft 365 Copilot, and other agent platforms.
Support for MCP and A2A
According to Microsoft, the Agent Framework combines and extends the concepts of the Semantic Kernel (a framework for integrating AI capabilities into applications) and AutoGen (a tool for automating agent-based workflows) projects. For developers who have previously worked with these, the transition should be seamless. The framework supports both the Model Context Protocol (MCP – a protocol for standardizing model interactions) and the Agent2Agent (A2A – a protocol for agent communication) protocol, is based on an OpenAPI-first design (prioritizing the use of the OpenAPI specification for APIs), and is cloud-independent.
According to Microsoft, the Agent Framework enables the following functions:
- Agents can dynamically discover and call external tools or data servers provided through the Model Context Protocol (MCP), which standardizes communication with external systems.
- Agents can collaborate across different runtime environments using structured messaging controlled by the Agent2Agent (A2A) protocol, which facilitates interactions among agents.
- Any REST API with an OpenAPI specification—a widely-used standard for describing APIs—can be immediately imported as a callable tool.
- Agents can run in containers (isolated software environments), on-premises (within an organization's own infrastructure), or across multiple clouds, making them portable across different environments.
The Agent Framework supports various orchestration patterns and includes an extension package for experimental features. It offers flexible memory modules compatible with services like Redis, Pinecone, Qdrant, Weaviate, Elasticsearch, Postgres, or custom storage.
Integrated connectors support Azure AI Foundry, Microsoft Graph, Microsoft Fabric, SharePoint, Oracle, Amazon Bedrock, MongoDB, and SaaS systems via Azure Logic Apps.
0 Response to "Microsoft has launched the Microsoft Agent Framework to simplify the creation of advanced AI agent workflows."
Post a Comment