Microsoft announces over 50 AI tools to build the ‘agentic web’ at Build 2025

Microsoft launched a comprehensive strategy to place itself in the center of what it calls “Web Open Web” at the annual construction conference this morning, as it presented dozens of tools and platforms of artificial intelligence designed to help developers create independent systems that can make decisions and completion tasks with limited human intervention.

RedMond company was provided. Technology GIANT more than 50 advertisements stretching on the entire product portfolio, from GitHub, Azure to Windows and Microsoft 365, all focused on the advancement of AI’s agent techniques that can work independently or cooperative to solve complex work problems.

“We have entered into the era of artificial intelligence agents,” said Frank Show, the chief communications official at Microsoft, in a blog post coinciding with the creation ads. “Thanks to the leading developments in thinking and memory, artificial intelligence models have now become more capable and effective, and we see how artificial intelligence systems can help problems solve new ways.”

How to transform artificial intelligence agents to develop software through independent capabilities

The “Web Web” concept moves beyond the artificial intelligence aides today. While the current artificial intelligence tools respond mainly to human questions and orders, agents start actively, make decisions independently, coordinate with other artificial intelligence systems, and complete complex workflow with the minimum human supervision. This is a fundamental shift in how artificial intelligence systems work and interact with both users and other technologies.

Kevin Scott, CTO of Microsoft, described this shift during the press conference as mainly changes how humans interact with technology: “Thinking of improvement will continue. We’ll see great progress there. But there is a handful of new things that should start to happen very quickly so that the agents are recipient of the most complex work.”

One of the missing items, according to silence, is the memory: “One of the things that is clearly lost at the present time in agents is the memory.” To address this, Microsoft offers many memory -related technologies, including structured RAG (the generation of retrieval), which helps artificial intelligence systems more precisely to summon information from large amounts of data.

“You will have a personal agent and a business agent, and the agent will have a full set of the employer’s information that belongs to each of you and the employer,” explained by Stephen Pathchi, CVP and his Microsoft’s technical colleague, during a presentation on agents.

Bathiche emphasized that this contextual awareness is very important to create agents “understands you well, weakens your place and what you want to do, and in the end you can understand so that you can click on less buttons at the end of the day.” This transformation of purely interactive intelligence to the continuous memory systems is one of the most deep aspects of the client revolution.

GitHub is evolving from completing the code to the independent developer experience

Microsoft puts GitHub, its famous developer platform, at the forefront of its agent strategy with the entry of an Github Copilot coding agent, which exceeds the code scraps to solve the programming tasks independently.

The new GitHub agent can now work as a member of software development teams, independently re -create the code, improve test coverage, stabilize defects, and even implement new features. For complex tasks, GitHub Copilot can cooperate with other agents at all stages of the program’s life cycle.

Microsoft is also a github coilot open conversation in the Visual Studio icon, allowing the developer community to contribute to its development. This reflects the double Microsoft approach to both the leadership of artificial intelligence with the adoption of the principles of open source.

“During the next few months, the capabilities that run with artificial intelligence materials will be part of the VS Open Source extensions, the same open source warehouse that drives the most popular software development tool,” the company explained in its announcement, with a focus on its commitment to transparency and societal innovation.

Multi -agent systems allow a complex workflow for business and automation of the operation

For companies looking to publish artificial intelligence agents, Microsoft has unveiled important updates for AZURE AI, a platform for developing and managing artificial intelligence applications and employees.

Ray Smith, Vice president of Artificial Intelligence Agents at Microsoft, highlighted the importance of multi -agent systems in an exclusive interview with Venturebeat: “The protest is multiple agents, correcting errors and digging in those multiple agents is essential, and this extends to what is essential in its Copilot studio to what will happen with AZURE

“It is extremely difficult to create a reliable process that you press in one agent. Dividing them into parts improves maintenance and makes building solutions easier, but it also greatly enhances reliability.”

AZURE AI Founder Agent, now available in general, allows developers to create Amnesty International agents at the level of institutions with the support of multi -agents and open protocols such as Agent2agent (A2A) and the McP Form Protocol (MCP). This enables organizations to organize multiple specialized factors to deal with complex tasks.

The potential of local artificial intelligence expands with energy transformations in the processing of customer devices

Although artificial intelligence based on a group of cuisine has dominated the main headlines, Microsoft is paying greatly towards the local AI on the device with many ads targeting developers who want to spread artificial intelligence directly on the user devices.

Windows Ai Foundry, a development of Windows Copilot, provides a unified platform for local artificial intelligence development on Windows. The Windows Ml operating system includes an operating time of artificial intelligence, and tools to prepare and improve models to spread the device.

The company announced: “Foundry Local will make it easy to run models, tools and agents of artificial intelligence directly on devices, whether Windows 11 or MacOS.” “Take advantage of ONNX, Foundry Local is designed for situations that users can save on using internet data, set privacy priorities and reduce costs.”

“We are very busy in an attempt to predict mainly and stay in the forefront. Most of our predictions are fulfilled within three or four months, which is a kind of madness, because I am used to predicting for a year or two, and then feels satisfied with that schedule. Now that is similar to that all the time, but this is fun.”

Challenges of Security and ID Department

With the spread of the agent’s use through institutions, Microsoft addresses the decisive need for safety, governance and compliance with many new capabilities designed to prevent what you call “expansion”.

Microsoft Entra, now in the inspection, the agents created by developers in Microsoft Copilot Studio or Azure Ai Foundry are automatically unique identities in Entra, which helps institutions to manage agents safely from the beginning and avoid “Sprawl” that may lead to blind sites, According to the announcement.

Microsoft also integrates data security controls and compliance with AI platforms, allowing developers to build Amnesty International solutions with safety and compliance features at the institution level. This includes controls of prevention of data loss for Microsoft 365 Copilot agents and new capabilities to detect sensitive data in artificial intelligence reactions.

Ray Smith advised the security technology technology teams: “Building solutions from A to Z is flexible, but then you have to add a lot of controls on these frameworks yourself. The beauty of Copilot Studio is that we offer you a framework of infrastructure with the Lifecycle administration and many of the capabilities of governance and observation that has been created.”

The scientific discovery platform shows how to convert artificial intelligence agents into the timelines for research and development

Perhaps one of the most ambitious applications for artificial intelligence agents announced in Build is Microsoft Discovery, a platform that aims to accelerate scientific research and development through industries from drugs to material science.

Jason Zander, CVP for Microsoft’s strategic missions and technologies, described in an exclusive interview with Venturebeat How this platform was used to discover non-PFAs for data centers in only 200 hours-a process that takes a traditional years.

“In our region, our data centers are huge for us because we are very excess,” Zandand said. “Using this framework, we managed to examine 367,000 potential candidates in only 200 hours. Then we took this to a partner who helped manufacture the results.”

Zander on how to represent this dramatic acceleration in the timelines for traditional research and development: “The descriptive point is, all those things that took, in some cases, years or even a decade.

Industry standards create an environmental system for operating factors via platforms

The Microsoft Center in Microsoft’s vision is to progress in open standards that enable the agent to operate with different platforms and services, with MCP Playing Playing Protocol (MCP) playing a particularly important role.

The company announced that it joined the MCP steering committee and has provided two new contributions to the MCP environmental system: updated licensing specifications and design to serve the MCP server.

Jay Bader, who leads the AI’s primary team in Microsoft, stressed the importance of openness and inter -operation: “Within Microsoft, all this relates to learning faster. Speed ​​is necessary because the world changes very quickly with new technologies and applications and competitors constantly emerging.”

Microsoft also presented NLWB, a new open project “that can play a role similar to the HTML for the Agentic Web”, allowing web sites to provide conversation interfaces for users with the model they choose and their own data.

Microsoft agent strategy is placed in the following computing model center

Microsoft ads in Build 2025 emphasizes the company’s all -in -the -intelligence agent as the following main computing model.

“The last time I was excited to be a developer or technically as I was now in the 1990s,” Kevin Scott said during the press conference. “One of the reasons why I feel this feeling in children’s stores with building blocks that even someone like me can understand. I can understand how each of these individual pieces works and how they consist together, and I can only play.”

Industry analysts note that Microsoft’s approach – combining Cloud and Edge AI, open standards with monopolistic technologies, developers tools with business applications – put the company as a central player in the emerging agent ecosystem.

For institutions customers, the immediate effect may be more evident in increasing the automation of complex workflow, the most intelligent responses of business events, and the ability to build custom agents that include knowledge and processes of the field.

With our transfer from a network of information to a network of agents, the Microsoft strategy reflects its previous approach to cloud computing – providing comprehensive tools, platforms and infrastructure with the progress of open standards simultaneously.

The question now is not whether artificial intelligence agents will transfer commercial operations, but how quickly institutions adapt to a world in which machines are not only responded to orders, but they expect needs, make decisions, and reshape how to do the work mainly.