Name: Axint
Author: Agentic Empire

Question 1

What is Axint?

Accepted Answer

Axint is an open-source compiler for Apple-native software. Author in TypeScript or Python today, use the preview `.axint` surface when you want the compact feature-shape format, and Axint lowers the definition into one IR before emitting production Swift for App Intents, SwiftUI views, WidgetKit widgets, and full apps. It also ships an MCP server so AI coding agents like Claude, Cursor, and Windsurf can compile and validate directly.

Question 2

Why does this matter right now?

Accepted Answer

Apple keeps expanding App Intents, SwiftUI, widgets, and system integrations across Siri, Shortcuts, Spotlight, and Apple Intelligence. That makes Apple-native integrations more important every release, but they are still expensive to build by hand. Axint gives teams and agents a faster path from TypeScript, Python, or preview `.axint` definitions to production Swift without skipping the native contract.

Question 3

What does Axint compile?

Accepted Answer

Axint v0.6.0 compiles Apple-native definitions into four Swift outputs: App Intents with @Parameter decorators and return-type-aware perform() signatures, SwiftUI View structs with @State, props, and #Preview blocks, WidgetKit stacks with TimelineEntry and TimelineProvider, and @main app scaffolds with Scene builders. Intent outputs can also produce Info.plist and entitlements fragments.

Question 4

How do I get started?

Accepted Answer

Install with `npm install -g @axint/compiler`, write a TypeScript file using any of the define*() APIs, and run `axint compile my-file.ts`. Or try it live — the playground on axint.ai compiles intents, views, and widgets as you type, with no install. The Axint Registry at registry.axint.ai has installable packages you can pull into your project with `axint add @axint/create-event`.

Question 5

Does Axint support Python?

Accepted Answer

Yes. The Python SDK ships on PyPI — install with `pip install axint`, then call `define_intent(...)`, `define_view(...)`, `define_widget(...)`, or `define_app(...)`. Python definitions parse into the same language-agnostic IR as TypeScript and compile through the same Axint pipeline, with native parsing, validation, and Swift generation.

Question 6

Is Axint free and open source?

Accepted Answer

Yes. Axint is released under the Apache 2.0 license with no CLA required. The source is at github.com/agenticempire/axint and accepts contributions. The Axint Registry is live at registry.axint.ai for browsing and publishing packages.

Question 7

What are Apple App Intents?

Accepted Answer

App Intents are Apple's framework for exposing your app's functionality to Siri, Shortcuts, Spotlight, and Apple Intelligence. Each intent declares a parameter schema and a `perform()` method that the system can invoke on behalf of the user. As Apple expands Siri and Apple Intelligence, App Intents remain one of the most important app-controlled integrations to invest in.

Question 8

Does Axint support Siri, Shortcuts, and Apple Intelligence?

Accepted Answer

Yes. Every Axint-generated intent is a standard Swift App Intent, which means it is automatically discoverable by Siri, Shortcuts, Spotlight, and Apple Intelligence on iOS 17+. As Apple expands Siri and Apple Intelligence, those same generated intents stay on the native system path. Views and widgets compile to standard SwiftUI and WidgetKit code that integrates with Apple system features directly.

Question 9

Can I use Axint with MCP (Model Context Protocol)?

Accepted Answer

Yes — Axint is MCP-native with install guides for Claude Code, Cursor, VS Code, Windsurf, Xcode, and more. The MCP server exposes 36 mcp tools + 5 prompts, including axint.compile, axint.validate, axint.swift.validate, axint.swift.fix, axint.schema.compile, and more. Your AI agent can draft a definition, compile it, validate the Swift, and open a PR without a human touching Xcode.

Question 10

What is the Axint Registry?

Accepted Answer

The Axint Registry at registry.axint.ai is an installable package registry for pre-built intent, view, widget, and app definitions. Browse live packages, install any with `axint add @namespace/package`, and publish your own with `axint publish`.

Question 11

Do I need to write Swift?

Accepted Answer

No. Axint is built so developers and AI coding agents can ship production Apple code without touching Swift. Use TypeScript or Python for the current production compile path, and use preview `.axint` when you want the compact declarative feature shape. The generated output is idiomatic and readable with full type fidelity for Int, Double, Float, Date, URL, and Measurement<UnitDuration>, so you can eject and hand-edit later if needed.

Question 12

Does Axint reduce token usage for AI coding agents?

Accepted Answer

Yes. The public benchmarks page currently measures token efficiency across the visible sample set and shows a 3.3x average compression, about 69% less definition overhead, using a fixed four-characters-per-token approximation. Treat it as token-efficiency proof for the measured examples, not a universal performance claim.

Question 13

How fast is the Axint compiler?

Accepted Answer

Fast enough for live editor and validation loops. Axint compiles a typical file to Swift, Info.plist, and entitlements quickly enough to power the live playground and interactive validation workflows without waiting on a local Xcode round-trip for every edit. The playground on axint.ai runs the same compiler path live for intents, views, and widgets.

Expose app actions and entities with contracts agents can verify.

Generate ordinary Swift

Validate the system-facing metadata

Treat model behavior as runtime evidence

Does every app need App Intents for Apple Intelligence?

Can Axint test Siri itself?

Put the contract into a real run.