إطار عمل وكيل Microsoft: توحيد الابتكار والاستعداد المؤسسي لتطبيقات الذكاء الاصطناعي الوكيلة

The Microsoft Agent Framework is a new open-source SDK and runtime designed to streamline the development, deployment, and management of sophisticated multi-agent systems. It aims to bridge the gap between experimental AI agent research and enterprise-grade production by combining the strengths of Semantic Kernel’s stable SDK and AutoGen’s innovative orchestration patterns. This framework empowers developers to build robust, scalable, and interoperable AI agents without compromising on innovation or enterprise readiness.

Why Microsoft Agent Framework?

• Addressing the Fragmentation: The Framework solves the issue of fragmented open-source AI agent frameworks, each with its own APIs and abstractions, hindering seamless transition from prototype to production.

• Unifying Strengths: It unifies the innovative agent orchestration patterns pioneered by AutoGen with the enterprise-ready features and stable SDK of Semantic Kernel. Developers no longer have to choose between experimentation and production.

Key Features and Capabilities

• Open Standards & Interoperability:

  • MCP (Model Context Protocol): Enables dynamic discovery and invocation of external tools and data servers, facilitating integration with a growing ecosystem of MCP-compliant services.
  • Agent-to-Agent (A2A): Allows agents to collaborate across runtimes using structured, protocol-driven messaging, enabling complex workflows across different frameworks and environments.
  • OpenAPI-first design: Simplifies the integration of REST APIs with OpenAPI specifications as callable tools.
  • Cloud-agnostic runtime: Agents can run in containers, on-premises, or across multiple clouds.

• Pipeline for Research to Production:

  • Supports sequential, concurrent, group chat, handoff, and magentic orchestration patterns.
  • Provides an extension package for experimental features, allowing advanced users to explore cutting-edge capabilities from Microsoft Research and the open-source community.

• Extensible by Design & Community-Driven:

  • Provides connectors to enterprise systems like Azure AI Foundry, Microsoft Graph, Microsoft Fabric, SharePoint, Oracle, Amazon Bedrock, and MongoDB.
  • Offers pluggable memory modules supporting Redis, Pinecone, Qdrant, Weaviate, Elasticsearch, Postgres, and custom stores.
  • Allows declarative agent definitions using YAML or JSON.

• Ready for Production:

  • Observability: Integrated OpenTelemetry support for visualizing agent actions, tool invocations, and orchestration steps.
  • Secure Cloud Hosting: Native integration with Azure AI Foundry for secure hosting with enterprise controls.
  • Security & Compliance: Integrates with Azure AI Content Safety, Entra ID authentication, and structured logging.
  • Long-Running Durability: Supports pausing, resuming, and recovering agent threads and workflows.
  • Human in the Loop: Allows marking tools as requiring human approval.
  • CI/CD Integration: Integrates with GitHub Actions and Azure DevOps pipelines.

Integration with Microsoft Ecosystem

• Microsoft 365 Agents SDK Integration: The Framework will integrate with the Microsoft 365 Agents SDK, enabling developers to build full-stack, multi-channel agents for Microsoft 365 Copilot, Teams, and the web.
• Azure AI Foundry Agent Service: A shared runtime with Azure AI Foundry Agent Service will provide a unified set of abstractions for creating, running, and scaling agents.

Migration Path for Existing Users

• Semantic Kernel Users: Migration involves replacing Kernel and plugin patterns with Agent and Tool abstractions. .NET developers will move to the Microsoft.Extensions.AI.* namespaces, while Python developers can install the full package or specific components.

• AutoGen Users: AssistantAgent maps to the new ChatAgent, FunctionTool wrappers migrate to the @ai_function decorator, and messaging is simplified with a unified ChatMessage type. Orchestration shifts to a graph-based Workflow API.

Customer Adoption and Momentum

Leading organizations like KPMG, Commerzbank, BMW, Fujitsu, Citrix, Fractal, TCS, Sitecore, NTT DATA, MTech Systems, TeamViewer, Weights & Biases, and Elastic are already testing and adopting Microsoft Agent Framework for a wide range of use cases, from automating audits and customer support to accelerating vehicle design cycles and enhancing IT support.

Conclusion

The Microsoft Agent Framework represents a significant step forward in the development of AI agents. By unifying innovation and enterprise readiness, providing robust features, and fostering a strong community, Microsoft is empowering developers to build the next generation of intelligent, autonomous applications. The framework’s integration with the Microsoft ecosystem and clear migration paths for existing users make it an attractive choice for organizations looking to leverage the power of AI agents.

Source: Microsoft Azure Dev Blogs