Apple WWDC 2026 / developer.apple.com Xcode 27 Ships Native AI Coding Agents: Claude, Gemini, and OpenAI Inside Apple's IDE
Apple announced Xcode 27 at WWDC 2026 on June 10 with native coding agents from Anthropic, Google, and OpenAI built directly into the IDE, alongside a local Neural Engine model for inline completions.
Apple announced Xcode 27 at WWDC 2026 on June 10, and the most significant change is not a UI refresh or a build system improvement. It is that Xcode now ships with native AI coding agents from Anthropic, Google, and OpenAI as first-class features, with no third-party extensions required.
Two Separate AI Systems in One IDE
Xcode 27 draws a clear line between two types of AI assistance.
Inline code completions run entirely on-device using Apple Silicon’s Neural Engine. No source code leaves your Mac for these suggestions. Apple handles the model and the inference, and it works whether or not you have an internet connection or an account with any external AI provider.
The coding agents — Claude from Anthropic, Gemini from Google, and OpenAI’s model — handle more complex, multi-step tasks. These do route requests to external servers, but only when you explicitly invoke an agent for a task. Apple made opt-in explicit: using a third-party agent requires developer confirmation, not just a default setting.
Both systems are available in Xcode 27 from launch. The inline completions require no configuration. For the cloud agents, you connect your Anthropic, Google, or OpenAI account and pay those providers at their standard API rates. Xcode itself remains free.
What the Agents Can Do
The coding agents in Xcode 27 go beyond chat and inline suggestions. Apple gave them access to the full development loop:
Agents can write code, but they can also run unit tests, inspect failures, and revise their work based on the test results. This means an agent can iterate on an implementation without you manually running tests between each attempt.
The new Device Hub gives agents access to physical devices and resizable simulators. An agent can deploy a build, tap through an interface, and check whether something looks right visually or behaves correctly, then report back and adjust the code.
Agents can also view SwiftUI previews, inspect build diagnostics, and use the Swift REPL to test expressions. The idea is that an agent working in Xcode 27 has the same feedback tools an experienced developer would reach for.
MCP and ACP for Extensibility
Xcode 27 uses two open protocols to keep the system extensible.
The Model Context Protocol (MCP) governs tool access. Apple ships a binary called mcpbridge that lets any MCP-compatible agent — Claude Code, Cursor, OpenAI Codex — connect directly to the live Xcode process. Through that connection they get structured access to diagnostics, symbol information, SwiftUI previews, and the Swift REPL in real time.
The Agent Client Protocol (ACP) handles which agents can be admitted beyond the three launch partners. If you want to connect Cursor or another agent that has implemented ACP, that path exists. Apple built the system to not be a closed three-vendor arrangement.
Apple’s Foundation Models API Also Got a Provider Layer
Separately from Xcode, Apple updated the Foundation Models framework at WWDC with a new LanguageModel protocol. The protocol is a public Swift interface that third-party cloud providers implement to expose a standard inference surface for iOS 27, macOS 27, iPadOS 27, watchOS 27, and visionOS 27 apps.
Anthropic shipped a Swift package that plugs Claude into this protocol. Google Gemini is also available. The practical effect: a developer can build an app against Apple’s on-device model, then swap in Claude or Gemini by updating a Swift Package Manager dependency. No changes to session logic or app code. Both providers are supported at launch.
This matters for app developers who want to use a more capable model for some features while keeping simpler queries on-device.
Platform Requirement
Xcode 27 requires Apple Silicon. Intel Macs are not supported. The on-device neural model is built around the Neural Engine, which Intel Macs do not have.
What This Means for the Broader AI Coding Tool Market
Apple choosing three separate AI providers rather than one — and using open protocols rather than a proprietary integration — is a meaningful signal. It puts Claude, Gemini, and OpenAI’s offering inside the IDE that most iOS and macOS developers already use, without requiring them to install a separate tool or change their workflow.
For the AI coding tool companies, native Xcode integration means reaching Apple platform developers who might never have installed Claude Code or Cursor on their own. The mcpbridge path also means existing tools can extend into the Xcode environment rather than being shut out.
Xcode 27 is available now in beta to Apple Developer Program members. It ships publicly with iOS 27, macOS Golden Gate, iPadOS 27, watchOS 27, and visionOS 27 this fall.
Sources:
The Weekly Diff
One email a week: every AI coding tool price change, plan restructure, and major release we verified, with sources. No filler.
Free. Unsubscribe anytime.