Last year, agents develop from experienced tools for basic corporate components. Of the simple robots for fast and responses to agents who act on your name, this shift refers to a new era of software design, where intelligence is not a long link to static interfaces or individual applications.
We have this transformation at Microsoft first. Azure AI Foundry It is now used by developing more than 70,000 companies and digital native companies, including Atomicwork, Epic, Fujitsu, Gainsight, H&R Block and LG Electronics, for designing, adapting and managing applications and agents AI. In just four months, more than 10,000 organizations have adopted our new service to build, deploy and scale of agency systems. More than 230,000 organizations, included 90% of the Fortune 500 they have already used Microsoft Braid studio.
Because the agents take over more sophisticated roles, they need access not only to different models and instruments, but also to OneTher. Therefore, we are determined to proceed open protocols such as the agent (A2A), soon to Azure AI Foundry and Copilot Studio, which will allow agents to cooperate across clouds, platforms and organizational slavery.
We compare with a wider pressure on the sectors for shared agents’ protocols, which is what we have always done: accepting open, support for developers in the real world and conversion to enterprise platform platforms. Our goal is simple: using developers Pro and for citizens to build this interoperation across clouds and framework.
We believe that Microsoft Copilot will seize every employee and act as “UI Pro AI” to teach with agents and agents -networks for this reason, acting and adapting via Bondarys. As customers expand, systems, Interoperability is not optional length. They want their agents to organize tasks that cover sellers, clouds and data silos. They want control, visibility and trust – the wing is locked.
A2A may allow the structured communication of the agent – exchange objectives, status management, evocation of actions and returning results safely and observable. Developers can use tools they know, such as semantic core or Langchain and still interoperate. Each travel travel to corporate warranties: Microsoft Entre, Mutual TLS, Azure AI Content Safety and full audit protocols. The Azure AI foundry is based on the default with confidence and how agents are growing more open and distributed, safety, compliance and responsibility remain first -class.
What we deliver
With A2A support:
- Azure AI Foundry Customers can create complex workflows with multiple agents that span internal copilots, partner tools and production infrastructure-possession of administration and SLA.
- Braid studio Agents will be able to safely trigger external agents, include those built with other platforms or hosted outside Microsoft.
- Businesses get a way to composible intelligent systems that spread across organizational and cloud borders.
- Microsoft’s contributions will speed up the development and acceptance of the open A2A protocol throughout the industry.
This is just one step we set out on the road. Since we do not do innovations such as autogen, semantic core, our contributions to the model contextual protocol (MCP) and our open models catalog, we will continue to develop a support platform Protocols, Models and Frames that most for developers and businesses. We see protocols such as A2A and MCP as important steps in the direction of realization of our vision for the agent future.
What will be next
Agency computer technology is not a trend – it is a basic shift. It changes how software is created, how it is decided and how value is created.
We joined the A2A working group on Github to contribute to specifications and tools. Soon the public view of A2A in Foundry and the Copilot Studio will arrive.
By supporting A2A and building on our open orchestration platform, we put the basis for the next generation of software – cooperation, observable and adaptive by design. The best agents will not live in one app or cloud; They will work in the work of work, spanning models, domains and ecosystems. We build this future with openness on Wednesday – Boot agents should not be islands and intelligence should work across the borders, as well as the world that serves.
We are starting
We introduced a new sample in the semantic core (available in Python), which shows how two local agents can work with A2A. In this example, agents work on planning travel itinerary and currency conversion processing, which represents trouble -free interoperability without the need for your own orchestration code.
Resources
(Tagstotranslate) transformation AI