Microsoft Build 2025: AI agents and building the open agentic web
Some highlights from Build 2025
At Microsoft Build 2025, one theme stood out above all: we are entering the era of AI agents. These are not just tools that assist, but intelligent systems capable of making decisions, orchestrating tasks, and interacting with users and systems in increasingly sophisticated ways. This year's announcements signal a major shift in how developers build, deploy, and secure applications—and how we imagine the future of the web itself.
AI Agents Are Becoming Central to Software Development
Across Microsoft’s platforms, AI agents are quickly becoming foundational. GitHub Copilot, which already supports over 15 million developers, is evolving into an asynchronous coding agent—a partner that goes beyond the in-editor experience to handle background tasks like code suggestions, debugging, and reviews.
Microsoft 365 Copilot is now used by over 230,000 organizations, including 90% of the Fortune 500. With new capabilities for multi-agent orchestration and Copilot Tuning, companies can now create specialized agents trained on their internal data and workflows. This empowers teams to automate domain-specific tasks, such as drafting legal documents or generating financial reports, all within the Microsoft 365 environment.
Azure AI Foundry is also expanding with the general availability of the Agent Service, which allows developers to coordinate multiple agents with built-in support for Semantic Kernel, AutoGen, and new protocols such as Agent-to-Agent (A2A) and the Model Context Protocol (MCP). These tools enable developers to build highly capable, enterprise-grade AI systems with security, observability, and trust built in from the start.
New Developer Platforms Empower Innovation at Scale
Microsoft continues to invest deeply in developer experience. The launch of Windows AI Foundry brings a unified platform for training and deploying AI models on Windows, with streamlined APIs for vision and language tasks. Developers can work with open-source LLMs locally or scale across the cloud, using the same toolchain.
Azure AI Foundry Models now offers more than 1,900 models, including the Grok 3 and Grok 3 Mini models from xAI, hosted directly by Microsoft. New tools like the Model Leaderboard help teams compare model performance across tasks, while the Model Router intelligently selects the best model for each query in real time.
In a significant move toward open-source collaboration, Microsoft is also open-sourcing GitHub Copilot Chat for Visual Studio Code. This reinforces GitHub’s role as the home for more than 150 million developers and highlights Microsoft’s commitment to open, collaborative, AI-powered software development.
Governance and Security for the Agentic Future
As AI agents become more prevalent, securing and managing them at scale becomes critical. Microsoft is addressing this challenge head-on with Microsoft Entra Agent ID, a new identity management solution that automatically assigns unique identities to agents built with Copilot Studio or Azure AI Foundry. This ensures that organizations can track and govern agents just like any other IT asset.
Integration with Microsoft Purview further extends security and compliance, offering tools to define risk parameters, run automated evaluations, and generate detailed reports on agent activity. These governance capabilities are essential for organizations looking to adopt AI responsibly and at scale.
Building the Open Agentic Web
One of the most forward-looking announcements at Build was Microsoft’s vision for an open agentic web—an internet where AI agents can discover, interpret, and interact with web content on behalf of users.
To support this, Microsoft is expanding support for the Model Context Protocol (MCP) across its platforms, and has joined the MCP Steering Committee alongside GitHub. Microsoft is contributing an updated authorization specification and a new MCP Server Registry, enabling secure, scalable agent interactions.
The introduction of NLWeb, a new open project, furthers this vision. Much like HTML enabled the modern web, NLWeb is designed to provide a semantic, conversational layer for AI agents to interact with websites using natural language. Every NLWeb endpoint can also act as an MCP server, giving website owners full control over how their content is accessed and used by agents.
Accelerating Scientific Discovery with Agentic AI
AI’s potential to transform science was another major theme at Build. Microsoft announced Microsoft Discovery, an extensible platform that supports research and development teams in accelerating their discovery processes. From drug development to sustainability, this platform is designed to bring agentic AI to some of the world’s most important challenges.
What’s new in data for AI and agents
Cosmos DB in Fabric (Preview)
Microsoft has introduced Cosmos DB as a NoSQL database option directly within Microsoft Fabric. This addition is particularly important because it handles semi-structured data like text documents, emails, and graphs that are essential for training AI systems. Developers can now deploy high-performance databases in just seconds while maintaining enterprise-grade reliability of 99.999% uptime. This technology has proven its scale at OpenAI, where Cosmos DB supports the trillions of database transactions needed to serve ChatGPT's 500 million weekly users.
Digital Twin Builder (Preview)
Microsoft has launched a new no-code and low-code tool called Digital Twin Builder that enables organizations to create virtual replicas of both physical and logical entities. Companies can use this tool to monitor machinery performance, track customer behavior, and optimize manufacturing processes. The tool is built on Fabric Real-Time Intelligence, which provides the foundation for scalable analytics and real-time monitoring capabilities.
Enhanced Copilot Experiences
Microsoft has expanded its AI-powered chat capabilities across multiple platforms. Power BI now features a full-screen Copilot experience that allows users to ask questions about their data across multiple reports and semantic models. Additionally, Fabric Data Agents serve as virtual business analysts that can connect to Microsoft Copilot Studio, enabling natural language conversations about organizational data. These features integrate directly with Microsoft Teams and Microsoft 365 Copilot, allowing users to access insights without leaving their normal workflow.
Major Azure Database Updates
SQL Server 2025 (Public Preview)
Microsoft has released SQL Server 2025 in public preview, transforming the traditional database into a vector database capable of supporting AI applications natively. The new version includes built-in vector search capabilities and integrates with popular AI frameworks including LangChain, Semantic Kernel, and Entity Framework Core. The preview has attracted over 3,400 applicants, with adoption rates occurring twice as fast as the previous SQL Server 2022 release.
PostgreSQL Enhancements
Microsoft has announced several improvements to its PostgreSQL offerings. The company has released a new PostgreSQL extension for Visual Studio Code that includes built-in GitHub Copilot integration, helping developers design schemas, write optimized queries, and troubleshoot issues using natural language. Microsoft has also made DiskANN generally available on Azure Database for PostgreSQL, providing one of the fastest vector indexing algorithms available. Additionally, the company has introduced semantic operators in preview that use large language models directly within PostgreSQL databases to provide deeper relational context from data.
Azure AI Foundry Integration
Microsoft has expanded integration between its database services and Azure AI Foundry. Developers can now use Azure Cosmos DB accounts to store and manage conversation threads between users and AI agents, allowing agents to maintain context across conversations. This functionality is generally available, with additional capabilities coming soon that will make Cosmos DB the first Azure database to power AI Foundry with real-time operational data. Microsoft has also introduced a public preview connection between Azure AI Foundry and Azure Databricks, enabling AI agents to use AI/BI Genie and run Databricks jobs for enhanced knowledge retrieval.
SAP Integration
Microsoft has deepened its partnership with SAP and Databricks to support the upcoming launch of SAP Business Data Cloud and SAP Databricks on Azure, expected in the third quarter of 2025. SAP Databricks will provide analytics capabilities specifically tailored to SAP workloads, allowing developers to build innovative solutions on top of core business data from SAP applications. When running on Azure, this service will benefit from Microsoft's native Azure Databricks service, providing robust scalability, performance, and security.
Microsoft Build 2025 made it clear that the future of AI is not just in chatbots or assistants—it’s in intelligent agents that act on our behalf, collaborate with other agents, and operate across cloud, desktop, and web environments.
Sources
If you’re interested in reading more, the links are below with more details:
https://blogs.microsoft.com/blog/2025/05/19/microsoft-build-2025-the-age-of-ai-agents-and-building-the-open-agentic-web/
https://azure.microsoft.com/en-us/blog/powering-the-next-ai-frontier-with-microsoft-fabric-and-the-azure-data-portfolio/