stclarke introduces Microsoft Agent Framework and Azure AI Foundry enhancements, offering developers unified tools to build, orchestrate, and govern enterprise-grade multi-agent AI systems with integrated observability and responsible AI features.

Introducing Microsoft Agent Framework: Streamlining Multi-Agent AI Systems with Azure AI Foundry

Author: stclarke
Source: Microsoft Blog

Overview

Microsoft has announced the public preview of the Microsoft Agent Framework, an open-source SDK and runtime that simplifies the orchestration of multi-agent AI systems. Alongside this release, Azure AI Foundry provides new capabilities that support developers with building, monitoring, and governing agentic solutions at enterprise scale.

Key Features

Unified Multi-Agent Platform

Microsoft Agent Framework brings together research and ready-for-enterprise foundations (merging AutoGen and Semantic Kernel).
Developers can experiment locally and deploy to Azure AI Foundry, benefiting from built-in observability, durability, and compliance.
OpenAPI-based API integration, Agent2Agent (A2A) runtime collaboration, and Model Context Protocol (MCP) support for dynamic tool connections.
Support for the latest multi-agent patterns, including Magentic One and orchestrated workflows.

Developer Experience & Ecosystem

Reduces context-switching and platform fragmentation.
Open-source connectors facilitate interoperability between Azure AI Foundry, Microsoft 365 Copilot, and external systems.
Strong focus on developer feedback and community contributions.

Enterprise Applications & Adoption

KPMG Clara AI uses the framework for audit automation, benefiting from enterprise-scale security and compliance.
Commerzbank pilots avatar-driven AI support, emphasizing productivity and operational reliability.
Citrix, TCS, Sitecore, and Elastic develop domain-specific solutions, from VDI productivity to content automation and enhanced search integrations.

Multi-Agent Workflows (Private Preview)

Foundry Agent Service introduces workflow orchestration, leveraging persistent state and context sharing for long-running tasks.
Visual authoring and debugging via VS Code Extension or Azure AI Foundry.
Supports complex enterprise scenarios such as customer onboarding and financial operations.

Observability & Standards

Contributions to OpenTelemetry standardize tracing and monitoring for agentic systems across frameworks.
Unified observability now available for agents built with Microsoft Agent Framework, LangChain, LangGraph, and OpenAI Agents SDK.

Voice Live API Availability

Voice Live API enables real-time, speech-to-speech agent solutions—integrating STT, TTS, avatars, and conversational features in a single API pipeline.
Customers (Capgemini, healow, Astra Tech, Agora) leverage this for voice-enabled customer service, HR, and education applications.

Responsible AI Features

Public preview launches for:
- Task adherence (details): Guiding agents to stay aligned with instructions.
- Prompt shields with spotlighting (details): Detect and mitigate prompt injection risks.
- PII detection: Identifying and managing sensitive data.
Built-in to Azure AI Foundry to foster trustworthy and compliant systems.

Customer Momentum & Use Cases

Over 70,000 organizations (from startups to large enterprises) are realizing ROI from Azure AI Foundry and Agent Framework-powered agentic AI solutions.
Testimonials and case studies from KPMG, Commerzbank, Citrix, TCS, Sitecore, and Elastic reflect the diversity and scalability of enterprise adoption.

Conclusion

Microsoft Agent Framework and Azure AI Foundry represent a unified, open, and responsible approach to agentic AI development. Developers can leverage cutting-edge tools, enterprise-grade governance, and broad ecosystem integration to accelerate the creation of scalable, trustworthy multi-agent systems.

Learn More:

References:

This post appeared first on “Microsoft News”. Read the entire article here