At Microsoft Build 2025, developers got a glimpse into how AI, security, and next-generation automation are transforming the way we work and create.
From the introduction of Windows AI Foundry, designed to streamline AI development, to MCP support enabling secure AI agents, Microsoft is making Windows 11 the ultimate platform for building intelligent applications.
Whether you’re an AI developer, IT administrator, or just passionate about technology, these innovations will make computing faster, smarter, and more connected than ever. Let’s dive into the most exciting updates announced at Build 2025 and explore how they’re shaping the future of AI on Windows.
Microsoft has introduced Windows AI Foundry, an evolution of the Windows Copilot Runtime, designed to streamline AI development on Windows 11. This platform integrates Windows ML, providing a built-in AI inferencing runtime that works across CPU, GPU, and NPU hardware from AMD, Intel, Nvidia, and Qualcomm.
Unleashing the Power of On-Device AI with Windows AI Foundry
The centerpiece of the Windows announcements is Windows AI Foundry, an evolution of the Windows Copilot Runtime. This unified platform aims to simplify the entire AI developer lifecycle, from choosing and optimizing models to deploying them seamlessly across both client and cloud environments.
Key Capabilities of Windows AI Foundry:
- Windows ML: The Foundation for AI Inferencing: Building upon the learnings from DirectML, Windows ML emerges as the built-in AI inferencing runtime on Windows. This empowers developers to bring their own AI models and deploy them efficiently across CPUs, GPUs, and NPUs from key silicon partners like AMD, Intel, NVIDIA, and Qualcomm.
- Simplified Deployment (Preview): Forget about packaging ML runtimes and hardware execution providers with your app! Windows ML will intelligently detect the target hardware, download the necessary components, and select the optimal execution provider based on your configuration. This promises a significantly smoother deployment process.
- Future-Proof Hardware Adaptation: The rapid evolution of AI hardware won’t leave your applications behind. Windows ML will automatically keep AI dependencies updated and adapt to new silicon while maintaining model accuracy and hardware compatibility, giving developers peace of mind.
- Powerful Tools for Performant Models (Preview): The AI Toolkit for VS Code will offer a robust suite of tools for preparing and optimizing models. This includes conversion to ONNX from PyTorch, quantization, optimization, compilation, and profiling. The goal is to simplify the creation of efficient models for Windows ML without the need for complex logic and multiple builds.
- Seamless Access to Open-Source Models with Foundry Local: Integrating with model catalogs like Ollama and NVIDIA NIMs, Foundry Local provides developers with quick and easy access to a rich ecosystem of ready-to-use open-source models directly on their Windows machines.
- Rich Catalog of OSS Models (Preview): Foundry Local will automatically detect your device’s hardware (CPU, GPU, NPU) and present a curated list of compatible models ready for experimentation.
- Command Line Interface (CLI) (Preview): Developers can leverage simple CLI commands like
foundry model listandfoundry model runto browse, test, and interact with models running locally. An SDK will also be available for seamless integration into applications.
- Ready-to-Use On-Device AI APIs for Windows Development: For Windows app developers looking to quickly integrate AI capabilities that run locally on Copilot+ PCs, Microsoft is introducing Windows AI APIs. These APIs are powered by inbox models shipped with Windows, offering a fast and easy entry point into on-device AI.
- LoRA for Phi Silica (Preview on Snapdragon X Series NPUs, coming soon to Intel and AMD): Low-Rank Adaptation (LoRA) enables efficient fine-tuning of models by updating only a small fraction of their parameters with custom data. This allows for improved performance on specific tasks without affecting the model’s core abilities. Developers can currently access LoRA for Phi Silica in Windows App SDK 1.8 Experimental 2.
- Knowledge Retrieval and Semantic Search (Private Preview): The new Semantic Search APIs empower developers to build intelligent search experiences within their applications using their own data. These APIs support both semantic search (understanding the meaning behind queries, including image search) and traditional lexical search. Running locally on all device types, these APIs offer excellent performance and privacy. On Copilot+ PCs, premium semantic capabilities are unlocked. Furthermore, these APIs support Retrieval-Augmented Generation (RAG), allowing developers to ground LLM outputs with their custom data.
- Existing APIs Move to Stable Release (Generally Available): Certain text and image APIs are now generally available for developers to integrate into their published applications, running locally and exclusively on Copilot+ PCs.
Enhancing App Discoverability and Agent Interactions
Recognizing the importance of app visibility and the emerging role of AI agents, Microsoft announced two key initiatives:
- Model Context Protocol (MCP) on Windows (Private Developer Preview with select partners in the coming months): MCP aims to establish a standardized framework for AI agents to seamlessly interact with native Windows applications. This will allow agents installed locally to leverage the specific functionalities of Windows apps, enhancing their skills and providing more value to users.
- MCP Registry for Windows: A secure and central hub for AI agents to discover available MCP servers on a Windows PC.
- MCP Servers for Windows: This will include core Windows functionalities like the File System, Windowing, and WSL as MCP Servers. Developers will also be able to wrap their app’s features as MCP servers, making them accessible to agents via the MCP Registry.
- Security is a paramount concern, and Microsoft has emphasized a robust security approach for MCP.
- App Actions on Windows (Available now in Windows SDK 10.0.26100.4188 or greater): This new capability, also available as built-in MCP servers, allows apps to expose their functionality to AI agents, increasing the discoverability of their features.
- App Actions APIs: Developers can use these APIs to define actions for their desired features and even consume actions from other relevant apps to offer complementary functionality, boosting user engagement.
- App Actions Testing Playground (Available via Microsoft Store): A dedicated environment for developers to test the functionality and user experience of their created actions.
Boosting Developer Productivity with Tooling Improvements
Microsoft also highlighted significant enhancements to popular Windows developer tools:
- WinGet Configuration (Generally Available in summer 2025): Simplifying development environment setup, users will be able to capture the current state of their device (including apps, packages, and tools from configured WinGet sources) into a single WinGet Configuration file. This allows for effortless replication of environments. WinGet Configuration will be updated to support Microsoft DSC V3, including app settings in the configuration file if the apps are DSC V3 enabled.
- Windows Subsystem for Linux (WSL) is now Open Source (Generally Available): This move empowers developers to access the WSL source code, contribute enhancements, participate in issue resolution, and learn collaboratively as a community.
- Advanced Windows Settings (Currently available in the Windows Insider Program): Addressing the challenge of customizing Windows with hidden settings, Advanced Windows Settings will provide a central location within the Windows Settings app to easily access and configure powerful, advanced options, such as enabling File Explorer with GitHub version control.
- Command Palette in PowerToys (Generally Available): The evolution of PowerToys Run, Command Palette, offers a centralized and customizable way for developers to access frequently used commands, apps, and workflows, reducing context switching and improving efficiency.
- Edit, the new command-line text editor on Windows (Currently open source, available in the Windows Insider Program in summer 2025): A new default command-line text editor called “Edit” will be included in Windows. Developers will be able to launch it by simply typing “edit” in the command line, allowing for quick file editing without leaving their current terminal workflow.
The Future of Windows Development is Intelligent and Integrated
By providing a unified AI platform with Windows AI Foundry, enhancing app discoverability through MCP and App Actions, and boosting developer productivity with significant tooling improvements, Microsoft is laying the groundwork for a more intelligent and integrated Windows experience. The focus on on-device AI capabilities, particularly on Copilot+ PCs, opens up exciting new possibilities for creating responsive, private, and innovative applications. Developers should keep a close eye on the preview releases and prepare to leverage these powerful new tools and features to shape the future of Windows.
While the promise of AI-driven experiences is compelling, one can’t help but feel a touch of nostalgia for the days when BUILD announcements also heralded more tangible updates to the Windows GUI and foundational developer APIs. We remember the excitement surrounding updates that directly translated into visibly better applications, more seamless integrations with connected devices, and noticeably optimized code that breathed new life into our machines. There was a certain magic in seeing how new UI paradigms and powerful APIs unlocked innovative features, and how performance gains made our everyday interactions smoother and more enjoyable.
As we embrace this new era of AI-centric development, there’s a hope that this focus doesn’t entirely eclipse the importance of those core areas. The desire for richer, more integrated user experiences with our ever-expanding ecosystem of devices, coupled with the need for applications that are not only intelligent but also performant and visually refined, remains strong. Perhaps the future will see a harmonious blend, where the power of AI is coupled with thoughtful GUI enhancements and robust APIs, leading to a new generation of Windows experiences that are both intelligent and intuitively delightful for users. Developers should keep a close eye on the preview releases and prepare to leverage these powerful new tools and features to shape the future of Windows, hopefully in a way that balances the cutting edge with the timeless appeal of a well-crafted and user-friendly operating system.
