Build for Apple & Android
Running natively on your devices is what sets IonClaw apart. Here is how to build and run the native Apple apps and the Android app from source.
Apple — iOS, tvOS & watchOS
The Apple project lives in apps/apple and targets iOS, tvOS, and watchOS from one shared core. The C++ engine compiles into an XCFramework, and the Xcode project is generated with XcodeGen.
You need macOS with Xcode, plus XcodeGen and CMake. Install the tools with brew install xcodegen cmake.
# clone the repository and enter it $ git clone https://github.com/ionclaw-org/ionclaw.git $ cd ionclaw # build the native xcframework for ios, tvos, and watchos # then generate the xcode project from project.yml $ make prepare-apple # open the generated project in xcode $ xed apps/apple → pick the iOS, tvOS, or watchOS scheme, then Run
The make prepare-apple target builds the XCFramework and runs XcodeGen for you. To run the steps separately, use make build-xcframework followed by make gen-apple.
Android
The Android project lives in apps/android, with the native app in app and a reusable ionclaw library in library. The C++ engine compiles to a native library for each Android ABI through the NDK.
You need Android Studio with the Android SDK and NDK, plus a JDK. Point ANDROID_NDK to your NDK installation before building.
# from the repo root, build the native libraries # then install and launch on a connected device $ make run-android → launches com.ionclaw.app on your device or emulator # or assemble the standalone ionclaw.aar for distribution $ make build-android-aar → outputs build/android-aar/ionclaw.aar
The make run-android target builds the native libraries, installs the debug build, and launches it. To work in the IDE, open apps/android/app in Android Studio and press Run.
Running models on device
Local GGUF models run on the device CPU, so the available RAM sets the limit. Check the recommended memory for each model size before you pick one for mobile.
View system requirements