Microsoft He has opened his event for developers Build 2025 With the presentation of its novelties in different sectors, showing a Support for open technologies. Thus, only a few hours after releasing the Windows subsystem for Linux as an open source, it has been clear that the Company’s commitment to AI agents For companies it goes through the Creation of an open agricultural networkwith which AI agents act in individual, organizational, equipment and business contexts making decisions and performing tasks in the name of users or entire organizations.
In addition to this, in Microsoft they emphasize that the entry of AI in the world of development is changing the way of writing, deploying and maintaining the code. Developers use it to free themselves from more tedious and routine work and focus on others considered more strategic. In addition, as the software development life cycle is transformed, the possibility of integrating new functions in developer support platforms arises, such as the novelties that the company has presented for Github, Azure AI Foundry or Windows himself.
In this sense, among the novelties Microsoft has brought to Build is a programming agent Github Copilot and updates at Github Models. With the first, Github Copilot goes from being an assistant in the editor to become an AI assistant with agricultural abilities, with the first asynchronous programming agent integrated in Github.
In addition, Github Models will now have prompts management, simplified evaluations and companies for companies. Microsoft has also released as Open Source Github Co -chat in vs Code, so that Github Copilot’s abilities.
Windows AI Foundryalready available, is another of the news announced in Build 2025 for developers. It offers a unified platform that supports the life cycle of the IA developer through training and inference. Through simple APIs of models for vision and language tasks, developers can manage and execute large models of Open Source language through local Foundry, or use their own model to convert it, adjust it and deploy it in the client and in the cloud.
In the Design, Personalization and Management Platform of Applications and Agents of AI AZURE AI Foundry, Microsoft has incorporated the models Grook 3 and Grok 3 Mini de xAIhosted and invoiced directly by the company. Thus, developers can choose from more than 1,900 models of the host by partners and Microsoft.
Model Leaderboard and Model Router are other of their novelties. The first is a tool that classifies the best performance models in different categories and tasks. The second is designed to choose an optimal model for a concrete consultation or task in real time.
Microsoft gives more capacity and security to AI agents
In Build 2025, Microsoft has presented new pre -designed agents, personalized agents construction blocks, multiagente capabilities and new models to facilitate the creation and deployment of agents safely. Azure AI Foundry Agent Service It is already available in a general way, so Microsoft offers new capabilities for the orchestration of specialized agents to manage complex tasks, including Semantic Kernel and Autogen in a single SDK, as well as support for A2A (Agent-to-Agent) and Model Context Protocol (MCP).
Azure Ai Foundry Observability Add functions for the integrated supervision of metric, quality, cost and safety. All with detailed monitoring in an optimized control panel. On the other hand, with Microsoft between agent, in the test version, the agents that developers create in Microsoft Copilot Studio or Azure Ai Foundry are automatically assigned unique identities in the board of directors, which facilitates the safe management of agents.
Applications and agents created with Foundry will also have the security of Purview data and their compliance controls. Foundry also offers improved government tools to establish risk parameters, execute automated evaluations and receive detailed reports.
Microsoft 365 Copilot Tuning It allows customers to use the data, workflows and processes of your company to train models and create agents using a low code approach. These agents can perform precise and specific tasks of a specific environment safely from the Microsoft 365 services environment. In addition, the multiagest orchestration of Copilot Studio connects several agents, which leads to the possibility of combining skills and addressing more complex tasks.
More opening for the agents network
Microsoft now offers a broad compatibility with the model context protocol (MSP) on its agents and frameworks platform. This is composed of Github, Copilot Studio, Dynamics 365, Azure Ai Foundry, Semantic Kernel and Windows 11.
In addition, Microsoft and Github have joined the MSP Steering Committee to collaborate in the evolution of this open protocol. For now, they have announced two contributions to the MCP ecosystem. The first is an updated authorization specification for people to use their trusted login methods to give agents and applications promoted by LLM access to data and services, such as personal storage units or subscription services.
The second is the design of a MSP Server Registration Service, which allows anyone who needs the implementation of public or private repositories, updated and centralized for MCP servers.
Microsoft has also announced another open project: Nrebin order to play a role similar to that of HTML in the agricultural network. This project facilitates that websites offer their users a conversational interface with the model of their choice and their own data, which will allow users to interact directly with the web content.
Each NLWEB Endpoint is also an MCP server, so websites can get, if they wish. that their contents can be easily discovered by AI agents, and be accessible to them.
Discovery, a platform to boost science with AI
Apart from the novelties for the agricultural AI for companies and for developers, Microsoft has also announced improvements and tools aimed at impulating science through artificial intelligence. The main one is the platform Microsoft Discoverydesigned so that researchers can transform the innovation process supported by agricultural AI.
This platform is specifically designed for R&D departments and research groups from several sectors, so that they can have tools to expedite the commercialization of new products, in addition to accelerating and expanding the research and development process throughout its phases to all scientists and researchers who intervene in the projects in which it is used.
