Xcode’s new AI agents can explore projects, write code, and fix errors

[kevin1]

Apple says its new agentic coding tools can now explore an app project, understand its structure, build the app, run tests, and fix errors automatically. The agents also reference Apple’s latest developer documentation to ensure correct API usage and best practices.

To make this work, Apple partnered closely with Anthropic and OpenAI, optimizing token usage and tool calling so agents run efficiently inside Xcode. The system is powered by Model Context Protocol (MCP), allowing Xcode to expose its tools — such as file management, previews, and documentation access — to compatible AI agents.

This approach also opens the door for third-party MCP-based agents to integrate directly with Xcode.

[mendez]

i’m cautiously optimistic but also ready for my build system to be “fixed” in ways i didn’t ask for.