Microsoft Build 2025: Developing An Open Agentic Web For The Age Of AI Agents

Remember when having virtual assistants was just a party trick? You could ask Siri to tell a joke or Alexa to play a song, while chatbots could set reminders or tell you the weather—that is, if you phrased it correctly.

Today, you’ve got ChatGPT summarizing reports, Browser Use managing tasks on your PC, and tools like GitHub Copilot suggesting entire blocks of code.

We barely noticed the shift—but AI has quietly moved from helper to handler, and today, it’s not just assisting but deciding, adapting, and acting.

This shift isn’t just happening in apps and browsers—but across the entire internet,

Well, that’s exactly what Microsoft Build 2025 was all about.


From May 19 until May 22, Microsoft is revealing its boldest vision yet: an “open agentic web” where AI agents don’t wait for your commands but understand the context, take initiative, and get things done across systems, tools, and even websites.

With sweeping updates across GitHub, Azure AI Foundry, Windows, Microsoft 365, and more, Microsoft is building an AI-powered world for intelligent agents to operate as digital teammates, not just tools.

So, here are some of the key announcements that were revealed at the event!
 

What Was Revealed At Microsoft Build 2025?

Microsoft Build 2025 introduced the concept of the “open agentic web”—an internet where AI agents act independently on behalf of users or organizations, powered by secure open standards. This is made possible by various collaborative AI agents, including:

GitHub

GitHub Copilot has evolved into a full-fledged asynchronous coding agent integrated within GitHub and Visual Studio Code. The platform now offers agent-based code review, deployment, debugging, and testing tools. GitHub Copilot currently supports over 15 million developers, while Microsoft 365 Copilot usage continues to rise in business and knowledge worker segments.

Microsoft is also open-sourcing GitHub Copilot Chat in VS Code, along with prompt management, lightweight evaluations, and enterprise-grade control features.
 

Azure AI Foundry

Azure AI Foundry added over 1,900 hosted models, including xAI’s Grok 3 and Grok 3 mini, enabling developers to choose the right models with secure data integration and governance tools.

The launch of Windows AI Foundry will enable developers to train, tune, and deploy open-source and proprietary LLMs locally or via cloud, with support for vision and language tasks via APIs.

Microsoft also revealed the Model Leaderboard and Model Router tools to evaluate and dynamically select the best-performing models for specific tasks in real time.

Moreover, with the Azure AI Foundry Agent Service generally available, developers can orchestrate multiple agents with specialized roles using Semantic Kernel, AutoGen, and Agent-to-Agent (A2A) capabilities. 
 

Microsoft Entra Agent ID

Next, the launch of Microsoft Entra Agent ID (in preview) will enable developers to give unique digital identities for every agent created in Azure AI Foundry or Copilot Studio to help reduce security blind spots. This will enable enterprises to leverage Microsoft 365 Copilot Tuning to customize agents using internal data and workflows, while multi-agent orchestration will allow multiple copilots to collaborate in executing complex, domain-specific tasks. 
 

NLWeb

Microsoft also introduced NLWeb, a new open project that acts like “HTML for agents,” by enabling websites to expose content and functionality via natural language interfaces. This will enable them to be better understood and used by AI agents.


Every NLWeb endpoint will also function as an MCP (Model Context Protocol) server—allowing discoverability, access control, and secure data permissions for agents interacting with web content. Microsoft announced full first-party support for MCP across GitHub, Windows, Azure AI Foundry, Copilot Studio, and Semantic Kernel, with an authorization specification and MCP server registry. 
 

Microsoft Discovery

Another reveal was a platform aimed at researchers, aiming to reimagine the scientific discovery process using AI agents. Microsoft Discovery promises to speed up time-to-market and accelerate and scale product innovation pipelines.

These open standards will enable AI agents to act on behalf of users while ensuring transparency, security, auditability, and interoperability between services and platforms.

Yet, this isn’t all that was revealed!
 

What Else Was Said At Microsoft Build 2025?

Microsoft announced that over 230,000 organizations, including 90% of the Fortune 500, have adopted Copilot Studio to build custom AI agents. This includes businesses like Fujitsu, NTT DATA, and Stanford Health Care, which are already piloting Microsoft’s AI agent orchestration tools to automate tasks, including sales prioritization and clinical preparation.

In a keynote speech, CEO of Microsoft Satya Nadella said, “You can ask questions and AI assistants give us answers. You can assign tasks to agents and have them execute them. Or work side-by-side with AI to complete jobs and projects. And you can mix and match all of these form factors.”

With these revelations (and more to come), do you think Microsoft’s vision for the open agentic web will redefine how we interact with AI agents?

Let us know in the comments below!