Apple is rolling out agentic coding in Xcode 26.3, enabling AI agents from Anthropic and OpenAI to access broader Xcode features, build projects, run tests, take image snapshots, consult developer documentation, and assist with code changes across the development workflow.

Developers enable agents in Xcode settings by linking AI provider accounts and choosing a model version, with options like GPT-5.2-codex or GPT-5.1-mini, and agents can be swapped within a project as tasks change.

A release candidate of Xcode 26.3 is available to developers now, with a full launch anticipated soon, following prior Xcode 26 releases that added AI model support.

The feature suite is described as agentic coding, integrating AI agents directly into the IDE to autonomously assist in app development.

The implementation is built on the Model Context Protocol (MCP), with Apple providing documentation to connect MCP-speaking agents to Xcode tools, and partners like Anthropic and OpenAI collaborating to optimize token usage and tool calls.

Agents can be any tool that uses MCP and can access current developer documentation, project structure, and Xcode features to perform tasks with transparency via a code-by-code change view and project transcript.

Setup involves downloading agents from Xcode settings, linking accounts, and using a transition workflow where the agent analyzes the project, consults docs, codes, builds, tests, and verifies functionality.

Interactions begin in a left-side prompt box where natural language requests define project goals or code changes, prompting agents to break tasks into steps and reference documentation before coding.

Code changes are highlighted in the editor, and a side project transcript shows under-the-hood activity to aid learning, debugging, and traceability.

Developers can add an agent with a single click and switch between agents mid-project as needed, with automatic updates when providers release new versions.

The current release cycle supports accessibility efforts, including a code-along workshop to demonstrate agentic coding in real time within Xcode.

Apple notes this shift is part of a broader push to integrate AI-assisted tooling across its ecosystem, with explicit emphasis on efficiency and reduced token usage for agents.