Introducing Microsoft Agent Framework

6 months ago 98

Microsoft Agent Framework, present successful nationalist preview, and caller capabilities successful Azure AI Foundry marque it easier for developers to build, observe, and govern multi-agent systems.

Today we’re announcing caller capabilities successful Azure AI Foundry that marque it easier for developers to build, observe, and govern multi-agent systems, portion helping organizations adjacent the spot spread successful AI.

As agentic AI adoption accelerates—eight successful 10 enterprises present usage immoderate signifier of agent-based AI, according to PwC1—the complexity of managing these systems is increasing. Developers look fragmented tooling, and organizations conflict to guarantee agents behave responsibly. Our latest updates to Azure AI Foundry code these challenges head-on. 

Introducing Microsoft Agent Framework (public preview) 

The Microsoft Agent Framework, present successful nationalist preview, is the open-source SDK and runtime that simplifies the orchestration of multi-agent systems. It converges AutoGen, a erstwhile Microsoft Research project, and the enterprise-ready foundations of Semantic Kernel into a unified, commercial-grade framework—bringing cutting-edge probe to developers.

With Microsoft Agent Framework, developers can: 

  • Experiment locally and past deploy to Azure AI Foundry with observability, durability, and compliance built in. 
  • Integrate immoderate API via OpenAPI, collaborate crossed runtimes with Agent2Agent (A2A), and link to tools dynamically utilizing Model Context Protocol (MCP). 
  • Use the latest multi-agent patterns similar Magentic One and orchestrate agents into Workflows. 
  • Reduce context-switching crossed tools and platforms. 
  • Build multi-agent systems connecting Azure AI Foundry, Microsoft 365 Copilot, and different cause platforms.

This model is designed to assistance developers enactment successful flow. According to an manufacture study2, 50% of developers suffer much than 10 hours per week owed to inefficiencies similar fragmented tools, highlighting the request for solutions that trim complexity and amended the developer experience.

One enactment utilizing Microsoft Agent Framework to trim friction is KPMG. KPMG’s translation began with KPMG Clara, its cloud-based astute audit level utilized connected each KPMG audit worldwide.

KPMG Clara AI is tightly aligned with the next-generation, open-source Microsoft Agent Framework, built connected the convergence of Semantic Kernel and AutoGen.

This means KPMG Clara AI tin link specialized agents to endeavor information and tools, portion benefiting from built-in safeguards and an open, extensible developer ecosystem. The framework’s open-source connectors let agents successful KPMG Clara AI to interoperate not lone with Azure AI Foundry, but besides with outer systems and connection models, making it imaginable to scale multi-agent collaboration crossed a global, regulated enterprise.

Foundry Agent Service and Microsoft Agent Framework link our agents to information and each other, and the governance and observability successful Azure AI Foundry supply what KPMG firms request to beryllium palmy successful a regulated industry.

— Sebastian Stöckle, Global Head of Audit Innovation and AI astatine KPMG International

We invitation developers to articulation america successful shaping the aboriginal of agentic AI by contributing codification and feedback to Microsoft Agent Framework. 

Multi-agent workflows (private preview) 

Building connected Microsoft Agent Framework, we’re extending these capabilities straight into the unreality with multi-agent workflows successful Foundry Agent Service. This caller diagnostic enables developers to orchestrate sophisticated, multi-step concern processes utilizing a structured, stateful workflow layer. 

With multi-agent workflows, teams can: 

  • Coordinate aggregate agents crossed long-running tasks with persistent authorities and discourse sharing. 
  • Automate analyzable endeavor scenarios specified arsenic lawsuit onboarding, fiscal transaction processing, and proviso concatenation automation. 
  • Leverage built-in mistake handling, retries, and betterment to amended reliability astatine scale. 

Workflows tin beryllium authored and debugged visually done the VS Code Extension oregon Azure AI Foundry, past deployed, tested, and managed successful Foundry alongside existing solutions. 

Azure A I Foundry interface showing a sequential workflow.

Several customers are presently experimenting with this capability, and we look guardant to expanding to much customers successful the coming weeks.

We’re besides announcing enhancements to multi-agent observability, with contributions to OpenTelemetry that assistance standardize tracing and telemetry for agentic systems.

This gives teams deeper visibility into cause workflows, instrumentality telephone invocations, and collaboration—critical for debugging, optimization, and compliance. We made the supra enhancements to OpenTelemetry successful collaboration with Outshift, Cisco’s incubation engine. 

With the supra enhancements, Azure AI Foundry present provides unified observability for agents built with aggregate frameworks, including Microsoft Agent Framework and others similar LangChain, LangGraph, and OpenAI Agents SDK.

Voice Live API successful Azure AI Foundry present mostly available 

Mutli-agent workflows are progressively initiated done dependable inputs and culminate successful dependable outputs. We’re excited to denote the general availability of Voice Live API, which empowers developers and enterprises to physique scalable, production-ready dependable AI agents. Voice Live API is simply a unified, real-time speech-to-speech interface that integrates speech-to-text (STT), generative AI models, text-to-speech (TTS), avatar, and conversational enhancement features into a single, low-latency pipeline. 

Organizations specified arsenic Capgemini, healow, Astra Tech, and Agora are leveraging Voice Live API to physique lawsuit work agents, acquisition tutors, HR assistants, and multilingual agents. Voice Live API is transforming however developers physique dependable AI agents by providing an integrated, scalable, and businesslike solution. 

Responsible AI capabilities nationalist preview 

Building connected advancements successful cause observability and model integration, it’s arsenic important to guarantee that AI systems run responsibly and securely—especially arsenic they go much profoundly embedded successful captious endeavor workflows. 

According to McKinsey’s 2025 Global AI Trust Survey3, the number 1 obstruction to AI adoption is deficiency of governance and risk-management tools. That’s wherefore we’re putting the pursuing liable AI features successful nationalist preview successful the coming weeks:

These capabilities are built into Azure AI Foundry, helping organizations physique with assurance and comply with interior and outer standards. 

Customer momentum 

Azure AI Foundry solutions are helping implicit 70,000 organizations worldwide—from integer natives to endeavor companies—transform AI innovation into applicable results. For example: 

Commerzbank: Commerzbank is piloting Microsoft Agent Framework to powerfulness avatar-driven lawsuit support, enabling much natural, accessible, and compliant lawsuit interactions.

The caller Microsoft Agent Framework simplifies coding, reduces efforts and afloat supports MCP for agentic solutions. We are truly looking guardant to the productive usage of container-based Azure AI Foundry agents, which importantly reduces workload successful IT operations.

— Gerald Ertl, Managing Director/Head of Digital Banking Solutions, Commerzbank AG

Citrix: Citrix is exploring however they tin usage agentic AI wrong virtual desktop infrastructure (VDI) environments to amended endeavor productivity and efficiency.

Citrix has ever embraced flexible ways of moving arsenic the person successful unafraid work. As we determination into a satellite wherever agentic AI works side-by-side with us, we’re excited to alteration that besides wrong workspaces that our customers already usage each day. Microsoft’s Agent Framework brings a modern, developer-first attack to gathering agents. With enactment for cardinal APIs and languages, and autochthonal adoption of emerging protocols for instrumentality calling and observability, it enables intuitive improvement of agents connected Azure AI Foundry without compromising developer control. We are anxious to leverage the model to present connected our imaginativeness – enterprise-scale, production-ready AI agents for our customers.

— George Tsolis, Distinguished Engineer, Citrix

TCS: Tata Consultancy Services is actively gathering a multi-agent signifier connected the Microsoft Agent Framework, with respective initiatives underway that showcase their strategical concern and method depth, including agentic solutions for finance, IT operations, and retail.

Adopting Microsoft Agent Framework is not conscionable a technological advancement, but a bold measurement towards reimagining manufacture worth chains. By harnessing Agentic AI and Frontier models, we alteration our teams to physique flexible, scalable, enterprise-grade solutions that alteration workflows and present worth crossed platforms. True enactment is astir empowering innovation, embracing change, and fostering an situation wherever agility and collaboration thrust excellence. 

— Girish Phadke, Head, Microsoft Azure Practice, TCS

Sitecore: Sitecore is processing a solution that enables marketers to interact seamlessly with the level by automating tasks crossed the full contented proviso chain—from creating and managing web experiences to handling integer assets—using intelligent agents.

By partnering with Microsoft to leverage its caller Microsoft Agent Framework, Sitecore tin bring unneurotic the champion of some worlds: the flexibility to powerfulness afloat non-deterministic agentic orchestrations and the reliability to tally much deterministic, repeatable agents. At the aforesaid time, we payment from Microsoft’s enterprise-grade observability and telemetry, ensuring that these orchestrations are not lone almighty but besides secure, measurable, and production-ready.

— Mo Cherif, VP of AI, Sitecore

Elastic: Elasticsearch supports a autochthonal connector to Microsoft Agent Framework, enabling developers to seamlessly integrate endeavor information into intelligent agents and workflows.

Elasticsearch is the discourse engineering level and vector database of prime for organizations to store and hunt their astir invaluable operational and concern data. With the caller Microsoft Agent Framework connector, developers tin present bring the astir applicable organizational discourse straight into intelligent agents and multi-agent workflows. This makes it easier than ever to physique production-ready AI solutions that harvester the reasoning powerfulness of agents with the velocity and standard of Elasticsearch.

— Steve Kearns, General Manager Search Solutions, Elastic

A trusted cause mill for developers 

Azure AI Foundry is much than a platform—it’s a trusted cause factory for developers and enterprises. Whether you’re a CIO looking to standard AI responsibly, a information designer focused connected governance, oregon a developer gathering the adjacent procreation of intelligent agents, Azure AI Foundry provides the tools, frameworks, and spot you need. 

Microsoft stands retired successful the AI scenery with its committedness to open standards, interoperability, and liable AI. The Microsoft Agent Framework, present successful nationalist preview, is simply a unified, enterprise-grade model that integrates cutting-edge probe and allows developers to seamlessly orchestrate multi-agent systems with built-in observability, durability, and compliance.

Unlike different solutions, our model supports integration with immoderate API via OpenAPI, collaboration crossed runtimes with Agent2Agent (A2A), and dynamic instrumentality connections utilizing MCP. This ensures developers tin trim context-switching and enactment successful flow, accelerating innovation.

The open-source quality of the model invites developers to lend and signifier the aboriginal of agentic AI, making it a genuinely collaborative and forward-thinking platform. With Microsoft, organizations tin spot that their AI systems volition beryllium powerful, efficient, responsible, and secure, addressing the apical barriers to AI adoption identified successful McKinsey’s 2025 Global AI Trust Survey.

1 PwC’s AI Agent Survey.

2 AI adoption is rising, but friction persists.

3 Insights connected liable AI from the Global AI Trust Maturity Survey.

Read Entire Article