Are these AI features truly private?

Yes, by using the FoundationModels framework, all data processing occurs on the device's local hardware, ensuring data never leaves the device and supports offline use.

What is the FoundationModels skill for Claude Code?

It is a specialized set of guidelines and code patterns that help Claude assist developers in integrating Apple's native, on-device language models into iOS applications.

What is the context window limit for on-device models?

On-device models typically have a limit of 4,096 tokens, which covers the combined total of instructions, user prompts, and the generated response.

How does the @Generable macro improve AI integration?

The @Generable macro allows developers to define Swift structures that the AI populates directly, providing type-safe, structured output instead of unpredictable raw text.

Which iOS version is required for these features?

The patterns provided in this skill target iOS 26 and later, which includes the necessary system frameworks for Apple Intelligence on-device models.

Apple FoundationModels On-Device

Name: Apple FoundationModels On-Device
Author: xu-xiang

byxu-xiang

•

323

•

Mobile Development

Integrates Apple’s on-device FoundationModels framework for secure, private, and offline LLM capabilities in iOS applications.

This skill provides comprehensive implementation patterns for leveraging Apple's native FoundationModels framework on iOS 26+. It enables developers to integrate on-device text generation, structured data extraction using the @Generable macro, and custom tool calling while maintaining total user privacy. By focusing on Apple Intelligence, it covers essential workflows like model availability checking, session management within hardware constraints, and real-time snapshot streaming for responsive, AI-driven user interfaces.

Características Principales

01Custom tool calling implementation for domain-specific AI actions

02On-device LLM session management and availability checking

03Real-time snapshot streaming for structured UI updates

04323 GitHub stars

05Privacy-first AI implementation with full offline support

06Structured output generation using the @Generable Swift macro

Casos de Uso

01Building privacy-preserving AI assistants within native iOS applications

02Implementing offline text summarization and generative features for mobile users

03Extracting structured data from natural language inputs without cloud dependencies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xu-xiang/everything-claude-code-zh foundation-models-on-device

For use in Claude.ai and ChatGPT

Características Principales

01Custom tool calling implementation for domain-specific AI actions

02On-device LLM session management and availability checking

03Real-time snapshot streaming for structured UI updates

04323 GitHub stars

05Privacy-first AI implementation with full offline support

06Structured output generation using the @Generable Swift macro

Casos de Uso

01Building privacy-preserving AI assistants within native iOS applications

02Implementing offline text summarization and generative features for mobile users

03Extracting structured data from natural language inputs without cloud dependencies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xu-xiang/everything-claude-code-zh foundation-models-on-device

For use in Claude.ai and ChatGPT