Microsoft Boosts Azure AI Foundry with Multi-Agent Enhancements
Microsoft is bolstering its AI capabilities with significant updates to its Azure AI Foundry. The tech giant is enhancing multi-agent observability and adding multi-agent workflows to the cloud, while also promoting interoperability with external systems.
Microsoft is contributing to OpenTelemetry to improve multi-agent observability, providing a unified view of agent activities. Additionally, it's introducing new multimodal AI models and a Gemini framework within Azure AI Foundry, which might be related to agent orchestration across tools and data sources.
The new framework supports Agent2Agent (A2A) and Model Context Protocol (MCP) for seamless communication between agents and systems. It combines AutoGen and Semantic Kernel, offering an open-source SDK and runtime for multi-agent orchestration. KPMG is already leveraging this framework to interoperate with Azure AI Foundry and external systems via its Clara AI smart audit platform.
Microsoft is also releasing new responsible AI features in public preview, including task adherence, prompt shields, and PII detection. These features aim to ensure agents behave as intended and protect sensitive information. Furthermore, Microsoft's Magentic-One can be used to orchestrate agents into workflows using the Gemini framework. Developers can experiment locally and deploy agents to Azure AI Foundry with built-in observability and compliance.
Microsoft's enhancements to Azure AI Foundry, including improved observability, multi-agent workflows, and responsible AI features, demonstrate its commitment to advancing AI capabilities. The new framework and tools simplify AI agent integration and workflow orchestration, making it easier for developers to create and manage AI agents.
