Microsoft unveiled preview APIs for the Windows Copilot Runtime (WCR) after a 9-month wait. This is a quick look at the initial experience and how to begin leveraging these new capabilities.
The release of Windows App SDK 1.7 Experimental 3 has made experimentation possible.
Requirements
The Windows Copilot Runtime (WCR) APIs require a Copilot+ PC. The documentation is inconsistent regarding whether this works only on Snapdragon X-based Copilot+ PCs or also on Intel Lunar Lake or AMD Zen 5-based PCs. It’s important to note that limitations tied to Snapdragon X may eventually change.
The development PC needs to be enrolled in the Windows 11 Insider Preview Program and running the latest Dev channel build (26120.3073 at the time of writing). This build is also available in the Build channel. The APIs access four of the small language models (SMLs) preinstalled on these PCs, and the newest versions—including related user features like Recall and Click to Do—are accessible through the Insider Program. This requirement will likely evolve over time.
(If you attempt to use an AI whose underlying model is absent, Windows will prompt you to check Windows Update, and it will download the model.)
Available APIs
The APIs in Windows App SDK 1.7 Experimental 3 are not exhaustive but offer a good starting point. These include:
Microsoft.Graphics.Imaging: Image APIs for image extraction, object selection, image scaling, and similar tasks.Microsoft.Windows.AI.ContentModeration: APIs aligned with Microsoft’s Responsible AI initiative, designed to filter potentially harmful content from prompts or generated content.Microsoft.Windows.AI.Generative: Core on-device generative AI APIs for both text and image creation.Microsoft.Windows.SemanticSearch: APIs offering semantic search capabilities, similar to those being tested in Windows 11 Search and File Explorer in the Dev channel, for pattern matching document and image files.Microsoft.Windows.Vision: APIs providing text and image object recognition capabilities within images.Microsoft.Windows.Workloads: The on-device AI functions within a new Workloads container, conceptually similar to the Windows Services host process (svhost).
The focus here is on generative text APIs.
The Dev Environment
Using the most recent stable version of Visual Studio 2022 is one option. You will need to add Windows App SDK 1.7 Experimental 3 support to your project.
Alternative Methods and Setup
Using Experimental Builds
Step 1: Enroll your development PC in the Windows 11 Insider Preview Program.
Step 2: Ensure you are running the latest Dev channel build (e.g., 26120.3073 or later).
Step 3: Install the Windows App SDK 1.7 Experimental 3. This provides the necessary APIs.
Visual Studio 2022 Setup
Step 1: Open Visual Studio 2022.
Step 2: Create a new project or open an existing one.
Step 3: Add support for Windows App SDK 1.7 Experimental 3 to your project.
Things to know
- The APIs require a Copilot+ PC.
- APIs access four small language models (SMLs) that come preinstalled on these PCs.
- You can attempt to access an AI for which the underlying model is not present, Windows will prompt you to check Windows Update and it will then download that for you.
The Windows Copilot Runtime provides a first glimpse at leveraging on-device AI capabilities within Windows applications, albeit with specific hardware and software requirements.