Xcode 27 Routes to Claude and Gemini Without Code Changes - WWDC 2026's Biggest Developer Announcement

Xcode 27 - The AI Coding System That Routes to Claude, Gemini, or OpenAI

The headline from the Platforms State of the Union is Xcode 27's new dual-engine AI coding system. It works in two layers: a local model running on the Mac's Neural Engine handles real-time Swift code suggestions with low latency and no cloud dependency. For heavier analysis tasks - full app simulation, test writing, complex refactors, visual inspection through live previews - Xcode 27 routes to a cloud model. The cloud model is your choice: Claude (Anthropic), Gemini (Google), or OpenAI's agents. Apple built this routing on a new Swift interface called the LanguageModel protocol - a public API that third-party cloud model providers implement to expose a common inference surface. Swapping between providers requires no code changes in your app or in Xcode's configuration - it is a dropdown in preferences.

The full Xcode 27 AI coding capability list from TechTimes: the coding agent can simulate entire apps, write and run tests, inspect visual changes through live previews, and operate the iOS Simulator through a new Device Hub. This is the most significant AI integration in Xcode's history - not a copilot suggestion bar, but a system that can autonomously simulate and test an app end-to-end. All of this ships in the developer beta released June 8 and is available now at developer.apple.com.

The LanguageModel Protocol - What It Means for Claude and Gemini Developers

The Foundation Models framework's new LanguageModel protocol is the most consequential new API for AI developers building on Apple platforms. It is a public Swift interface that any third-party cloud model provider can implement to expose a common inference surface across iOS 27, macOS Golden Gate, iPadOS 27, watchOS 27, and visionOS 27. Apple's own models implement the protocol. Claude and Gemini implement the protocol as confirmed launch partners.

What this means in practice: an iOS app developer can write their AI feature calls against the LanguageModel protocol once, and at runtime the user's or developer's configured model handles the inference - whether that is Apple's on-device model, Apple's Private Cloud Compute model, Claude, or Gemini. A user who prefers Claude can get Claude-powered features in apps that support the protocol. A developer who wants to offer multiple model choices to enterprise customers can do so without maintaining separate code paths for each provider.

This is a distribution event for both Anthropic and Google that goes beyond any previous enterprise partnership or developer integration. Every iOS and macOS developer who uses the LanguageModel protocol is potentially routing users to Claude or Gemini. Apple has 34 million registered developers. Even a small fraction adopting Siri AI integrations through the protocol represents massive new demand.

Apple Foundation Models - Free Under 2M Downloads

Apple is offering free Private Cloud Compute access to Apple Foundation Models for developers with fewer than two million first-time App Store downloads. MacRumors confirmed this as the centerpiece of the Platforms State of the Union. The threshold is meaningful - two million first-time downloads covers the vast majority of iOS developers by number (the App Store has hundreds of thousands of apps, most of which never reach two million downloads), while still requiring large-scale apps to pay for the infrastructure they use.

What Apple Foundation Models can do in the developer beta: text generation, summarization, and classification running either on-device (Neural Engine) or on Apple's Private Cloud Compute servers via the same API. Image input support ships alongside the framework. The Dynamic Profiles system allows multi-agent workflow construction - an orchestrator model managing sub-agents - within the same Swift API surface. This is the foundation for Apple's vision of AI that runs entirely within Apple's privacy boundary for most use cases, falling back to Claude or Gemini only for queries that exceed on-device capability.

SiriKit Deprecated - What Developers Need to Do

Apple gave SiriKit a formal deprecation notice at WWDC 2026. The support window is two to three years - roughly the iOS 29 timeframe, expected around fall 2028. SiriKit will continue to work until then; it will not break immediately. But it will not receive new features, and apps that rely on SiriKit Intents for Siri voice integration should plan migration to the new App Intents framework.

The migration path: App Intents (introduced iOS 16, significantly expanded in iOS 17 and 18) is SiriKit's replacement. It is a cleaner Swift API, supports Siri AI's new personal context capabilities (accessing user emails, messages, and files to complete actions), and is the integration surface for the rebuilt Siri. Apps that have already migrated to App Intents are positioned for the full Siri AI feature set from iOS 27 onwards. Apps still on SiriKit will continue to work but will not benefit from Siri AI's contextual capabilities.

Frequently Asked Questions