Agentic programming in Xcode: more than just a code autocomplete

Xcode 27 fundamentally changes the way developers work with artificial intelligence. While previous versions primarily offered line-by-line code completion, the new release brings full-fledged agent assistants from three different providers. They handle not only code writing but also complex tasks such as automated test generation, interactive application architecture planning, or independent idea testing in the new Device Hub — an environment where AI agents verify code functionality on simulated devices.

Crucially, developers can choose a model based on their needs: ChatGPT (OpenAI) excels at creative problem solving, Claude (Anthropic) shines in large codebases, and Gemini (Google) brings deep integration with the Apple ecosystem thanks to the new Foundation Models architecture. Susan Prescott, Apple's Vice President of Worldwide Developer Relations, said: "Developers are at the heart of the Apple ecosystem, and our goal is to provide them with the best possible tools."

Core AI: large language models running directly on Apple silicon

For developers who want to keep everything under control and avoid sending data to the cloud, Apple introduced the Core AI framework. It allows deploying full-fledged large language models (LLMs) directly on Apple Silicon devices and leveraging their Neural Engine for local inference. This means faster response times, lower latency, and most importantly, no data leaves the device — which is crucial for companies handling sensitive information or for applications subject to the European GDPR regulation.

For smaller developers and startups, Apple has prepared an attractive incentive: members of the App Store Small Business Program (studios with fewer than 2 million downloads) get free access to the new generation of Apple Intelligence models running on Private Cloud Compute — Apple's secure cloud environment.

Google Gemini and Apple: a partnership that surprised

One of the most interesting announcements of WWDC is the deep collaboration between Google and Apple. Apple opened its Foundation Models framework to third-party developers through a new public LanguageModel protocol. Google, as the first major partner, integrated Gemini models into it — via the Firebase Apple SDK and the Firebase AI Logic service.

What does this mean in practice? Developers use a single API through which they can switch between Apple's local model and Google's cloud-based Gemini models depending on what their application needs at the moment. "Developers gain seamless access to Gemini models to build dynamic experiences for their end users while increasing their own development velocity," stated Nicholas McNamara and Thevi Sundaralingam from Google on the company's official blog.

For individual developers, Gemini in Xcode is available through a personal API key from Google AI Studio, which offers both free and paid tiers. Enterprise teams can use the Gemini Enterprise Agent Platform with dedicated quotas and enterprise-grade data protection parameters.

What it means for Czech developers and companies

For the Czech development scene, this is a significant moment. Xcode 27 and all beta versions are available immediately — simply sign up for the Apple Developer Program, which costs $99 per year (approximately 2,300 CZK). Czech developers building apps for iOS and macOS thus gain access to the same tools as Silicon Valley giants.

From an EU regulatory perspective, it's significant that Apple gives developers a choice between cloud and local inference. The Core AI framework running on the Neural Engine means that sensitive data does not have to leave the device, making it easier to comply with GDPR and potentially with the requirements of the EU AI Act. At the same time, integration with Firebase AI Logic from Google means developers don't have to build and manage their own backend servers for AI features — everything runs on Google's infrastructure, which meets European standards.

The question of the Czech language is also interesting. While ChatGPT and Claude handle Czech very well and Gemini also manages, Apple Intelligence and Siri AI do not yet natively support Czech. While this doesn't directly affect Xcode, it does impact how well AI agents will understand Czech-language comments and documentation. For Czech teams that document code in English, this won't be a problem — but smaller studios writing in Czech may hit limitations.

Swift 6.4, Liquid Glass, and gaming tools

Xcode 27 doesn't just bring AI. The update includes Swift 6.4 with improved concurrency support and a new Liquid Glass visual style — transparent interface elements with customizable effect intensity. For game developers, Apple prepared the Steam Asset Converter, which simplifies porting PC games to Apple devices, and Game Porting Toolkit 4, which uses AI agents to optimize games for the Metal API.

Comparison with the competition: Xcode catches up, but with grace

It's fair to say that Apple has been a bit late in the AI-assisted development space. While GitHub Copilot (Microsoft) and Cursor integrated AI into development tools more than two years ago, Apple is only now coming with an answer. But what an answer — instead of one model, three at once; instead of cloud-only, local inference too; and all of it free for small developers. For comparison: GitHub Copilot Max costs $100 per month, while Xcode 27 is part of a free IDE where you only pay for API calls to cloud models.

The "let developers choose their model" strategy is smart. No AI model is universally the best — Claude excels at large-scale refactoring tasks, ChatGPT at creative solutions, and Gemini benefits from deep integration into the Google ecosystem. The developer can choose based on the specific task at hand.