If you're building AI agents that need to generate Apple-native code, this gives you 36 tools that speak App Intents, SwiftUI, and WidgetKit. You define intents, views, and widgets in TypeScript or Python, and it compiles to Swift with all the boilerplate, validates against Apple's rules, and returns structured repair packets when something breaks. The MCP methods include compile, validate, repair, and schema operations. The premise is that general coding agents waste tokens regenerating Widget structs and TimelineProvider conformances. This narrows the contract to defineIntent() and defineWidget() calls, emits ordinary Swift you can inspect, then hands the agent a fix packet with ranked files and classified evidence when the next iteration needs guidance.
claude mcp add --transport stdio agenticempire-axint uvx axint