Noticia
Apple opens new on-device AI tools and agentic Xcode to app developers
Apple’s June 8 developer announcement adds new AI APIs, model choices and agent-assisted workflows for apps across iPhone, iPad, Mac and more.
Apple is giving developers a broader set of tools for building intelligent apps across its platforms. Announced on June 8, 2026, during WWDC26, the update combines new AI frameworks with a major expansion of Xcode’s agent-assisted development features.
The announcement is aimed at developers rather than ordinary iPhone owners, but it could shape how future apps handle text, images, voice commands and automated actions. Apple is also opening the door to models from other providers, giving development teams more choice over where AI processing takes place and which model powers a feature.
A single Swift API for more capable app intelligence
Apple’s Foundation Models framework is being expanded into a single native Swift API for app developers. According to Apple, the framework now supports more powerful on-device models, image input, server-based models and custom skills. That gives developers a common route for adding AI features without building a separate integration for every model or execution environment.
The framework can also connect apps to models from providers such as Anthropic and Google, as well as other providers that implement Apple’s new language model protocol. Apple says Dynamic Profiles will let developers change how a model interacts with an app without redesigning the entire integration.
For users, the practical result could be more capable features inside familiar apps. A notes app might interpret an image as well as text, a productivity tool could use a custom skill to structure information, and a creative app could choose between local processing and a server model depending on the task. Those examples describe the possibilities of the tools, not features Apple has promised in any specific third-party app.
More control over where AI runs
Apple is also introducing Core AI for developers who want to bring their own models to an app. The framework is designed to run custom models on Apple silicon and is optimized for the platform’s unified memory and Neural Engine. Apple describes it as a way to deploy full-scale large language models locally, which may help teams design features that work without sending every request to a remote service.
The announcement does not mean every AI feature will operate offline. Apple’s developer tools support both on-device and server models, so the final balance will depend on the app, the model and the developer’s privacy and performance choices. Users should therefore continue checking an app’s privacy information rather than assuming that an AI label automatically means local processing.
Apple is also lowering one barrier for smaller developers. Teams enrolled in the App Store Small Business Program with fewer than two million total first-time App Store downloads can access the next generation of Apple Foundation Models through Private Cloud Compute without a cloud API charge, according to Apple. That offer applies to eligible developers, not to consumers downloading an app.
Xcode 27 turns AI assistance into part of the workflow
The other major part of the announcement is Xcode 27. Apple is bringing coding agents from Anthropic, Google and OpenAI into the development environment, with interactive planning, multi-turn questions and a workspace that can display code changes and previews alongside the conversation.
Apple says the agents can write and run tests, experiment in Playgrounds, inspect visual changes in previews and interact with simulators through the new Device Hub. Developers can manage physical devices, resize simulators and test apps from one workspace. Xcode plug-ins can also connect outside tools through the Model Context Protocol and the Agent Client Protocol; Apple identifies GitHub and Figma as the first partners with seamless installation.
These features are designed to shorten the distance between an idea and a tested build, but they do not remove the need for human review. A coding agent can propose an implementation or generate tests, while the developer remains responsible for checking behavior, permissions, privacy, accessibility and release requirements.
Apps can become easier to find and use
Apple is updating App Intents so developers can connect app actions and content to Siri AI capabilities such as personal context understanding and onscreen awareness. If developers adopt the framework, users may be able to reach more app functions through natural-language requests instead of navigating through several menus.
That could make a meaningful difference on a phone, where small screens and frequent context switching often make complex workflows awkward. It also gives developers another reason to describe app actions clearly and expose them through Apple’s system frameworks rather than keeping every useful function hidden inside the app interface.
What is available now
Developer betas for iOS 27, iPadOS 27, macOS 27, watchOS 27, tvOS 27, visionOS 27 and Xcode 27 were listed as available through Apple’s developer channel from June 8. Xcode 27 is Apple-silicon-only, and Apple Intelligence features remain limited to supported regions.
For smartphone users, the important point is that these tools are an early foundation rather than an immediate promise of a finished feature in every app. The next visible changes will depend on developers adopting the frameworks, testing their models responsibly and shipping updates that explain what is processed locally, what is sent to a server and which permissions are required. Apple’s official announcement provides the full list of frameworks, tools and availability details.