What AI tools to use for iOS development

Issue #1020

AI coding assistants struggle with Swift. They suggest deprecated APIs. They miss SwiftUI performance issues. They write concurrency code that crashes at runtime.

The iOS community has built tools to fix this. Agent skills teach AI assistants modern patterns. MCP servers connect them to Xcode and Apple docs. CLI tools automate app distribution.

Agent Skills

Agent skills load specialized knowledge into AI assistants. When you ask Claude Code to review your SwiftUI code, the skill provides context about best practices and common mistakes. Skills load on demand, so they only use tokens when needed.

Antoine van der Lee

Antoine van der Lee maintains skills built from production experience at WeTransfer.

• SwiftUI Agent Skill — state management, composition, and performance
• Swift Concurrency Agent Skill — async/await, actors, and Swift 6 migration
• Core Data Agent Skill — persistence, thread safety, and migrations
• Swift Testing Agent Skill — testing macros and XCTest migration

Thomas Ricouard

Thomas Ricouard publishes a skills collection for recurring engineering tasks. It includes iOS Debugger Agent for build and debug workflows, SwiftUI Performance Audit for finding invalidation storms, SwiftUI View Refactor for breaking views into components, SwiftUI UI Patterns for navigation examples, SwiftUI Liquid Glass for iOS 26+ effects, Swift Concurrency Expert for actor isolation issues, and App Store Changelog for generating release notes.

Paul Hudson

Paul Hudson maintains skills focused on common AI mistakes.

• SwiftUI Agent Skill — deprecated APIs, accessibility, and edge cases
• SwiftData Agent Skill — models, queries, and iCloud sync
• Swift Concurrency Agent Skill — async/await, actors, and Sendable
• Swift Testing Agent Skill — test macros and parameterized tests

Finding More Skills

• skills.sh — directory for discovering agent skills, supports Claude Code, Cursor, and Copilot

Install any skill with npx skills add <owner/repo>.

CLI Tools

• App Store Connect CLI — automates app distribution, TestFlight, review submission, and metadata updates

MCP Servers

MCP servers extend AI assistants with external tools.

• GitHub MCP Server — browse code, manage issues and PRs, monitor Actions workflows
• XcodeBuildMCP — compile for simulators and devices, parse build output, capture logs
• String Catalog MCP — manage localization files, find missing translations, bulk translate

Xcode MCP

Xcode 26.3 introduces native support for agentic coding through the Model Context Protocol. External AI agents can now interact directly with Xcode’s build system, SwiftUI previews, and documentation search.

Open Xcode Settings with ⌘, and navigate to the Intelligence section. Toggle “Xcode Tools” to ON under Model Context Protocol. Xcode displays a visual indicator when external agents connect.

Connecting Claude Code

Run this command to register Xcode as an MCP server:

claude mcp add --transport stdio xcode -- xcrun mcpbridge

The mcpbridge binary acts as a translator between the MCP protocol and Xcode’s internal XPC layer. Xcode must be running with a project open. Once it is, mcpbridge auto-detects the Xcode process and connects without manual configuration.

Xcode exposes over 20 MCP tools across several categories:

Two tools stand out. RenderPreview captures screenshots of your SwiftUI previews, allowing agents to verify UI changes visually before committing code. DocumentationSearch queries Apple’s entire documentation corpus and WWDC transcripts using Squirrel MLX, an on-device semantic search system.

Documentation

• Sosumi AI — converts Apple docs to Markdown for AI access, covers APIs, HIG, and WWDC transcripts