Skip to main content
Apple's Xcode 26.3 with Claude and Codex integration for AI-powered auto-entitlement of APIs.

Editorial illustration for Apple adds Claude and Codex to Xcode 26.3, enabling auto‑entitlement for APIs

Apple adds Claude and Codex to Xcode 26.3, enabling...

Apple adds Claude and Codex to Xcode 26.3, enabling auto‑entitlement for APIs

2 min read

Apple’s latest Xcode update, version 26.3, slips two heavyweight language models—Anthropic’s Claude and OpenAI’s Codex—directly into the IDE. The move isn’t just a gimmick; it’s an attempt to give developers a “coding assistant” that can act on their behalf inside the same sandbox where the app is built. While the integration itself feels incremental, the real shift lies in how the assistants handle the platform’s security gatekeepers.

Previously, an external AI could suggest code that calls a protected API, but the developer still had to manually edit the project’s entitlement file—a step that often involved digging into binary formats and obscure settings. By embedding the models, Apple lets the assistant recognize when a capability is required and insert the correct entitlement without the programmer’s direct input. This capability, rare in today’s toolchains, promises to smooth out a traditionally clunky part of the iOS development workflow.

As Andrej Karpath notes, the result is a task that would otherwise be “very difficult to do” for an AI operating outside the development environment.

Agents can now automatically add entitlements to projects when needed to access protected APIs -- a task that would be "otherwise very difficult to do" for an AI operating outside the development environment and "dealing with binary file that it may not have the file format for." From Andrej Karpathy's tweet to LinkedIn certifications: The unstoppable rise of vibe coding Apple's announcement arrives at a crucial moment in the evolution of AI-assisted development. The term "vibe coding," coined by AI researcher Andrej Karpathy in early 2025, has transformed from a curiosity into a genuine cultural phenomenon that is reshaping how software gets built. LinkedIn announced last week that it will begin offering official certifications in AI coding skills, drawing on usage data from platforms like Lovable and Replit.

Apple's Xcode 26.3 now ships with Anthropic's Claude Agent and OpenAI's Codex built in. This integration lets the agents write code and, crucially, add entitlements automatically—a step that developers have traditionally handled manually. A bold shift.

Because the AI operates inside the IDE, it can manipulate binary files and grant access to protected APIs without external tooling. Yet the move raises questions: will developers trust an autonomous system to edit project metadata? While the feature promises to simplify a task described as “very difficult” for out‑of‑context AI, it also introduces a new layer of complexity around trust and verification.

The term “agentic coding” signals an emerging practice that remains controversial among the developer community. Apple positions the update as a push toward more autonomous development workflows, but it is unclear whether the benefits will outweigh the risks of unintended entitlement changes. As the release candidate rolls out, developers will be watching to see if the convenience translates into reliable, secure builds, or if additional safeguards become necessary.

Further Reading