Platforms & Requirements

Yukkai runs natively on macOS and iOS. Learn about supported platforms, minimum system requirements, and recommended hardware for the best experience.

Yukkai is a native app built with SwiftUI, available on macOS and iOS. It uses Swift Concurrency (async/await, actors) and takes full advantage of Apple hardware.

Supported Platforms

Platform	Minimum OS	Recommended
macOS	macOS 14 Sonoma or later	macOS 15 Sequoia or later
iOS	iOS 17 or later	iOS 18 or later

Mac Hardware Requirements

Minimum

Chip: Apple Silicon (M1 or later) — Yukkai is not supported on Intel Macs.
RAM: 8 GB unified memory.
Storage: 10 GB free space (more required for local models — see below).
Display: Any resolution supported by macOS.

Why Apple Silicon?

Yukkai leverages Apple Silicon's unified memory architecture and Neural Engine for:

Local model inference via Apple's MLX framework (GPU-accelerated on-device LLM inference).
On-device OCR via Apple Vision framework.
Fast screenshot capture via ScreenCaptureKit.
Secure Enclave for TouchID-protected vault encryption.

iOS Hardware Requirements

Device: iPhone 12 or later (A14 Bionic or newer).
RAM: 4 GB minimum, 6 GB+ recommended.
Storage: 5 GB free space (local models require significantly more).

Note: Some features are macOS-only (desktop automation, bash execution, IDE integration). iOS has a simplified tool set — see the iOS Limitations section below.

macOS Permissions

Yukkai requires several macOS permissions to function fully. You'll be prompted during onboarding:

Permission	Required For	How to Grant
Accessibility	Desktop automation (clicks, typing, window management)	System Settings → Privacy & Security → Accessibility
Screen Recording	Screenshots via ScreenCaptureKit	System Settings → Privacy & Security → Screen Recording
Full Disk Access	File read/write across the filesystem	System Settings → Privacy & Security → Full Disk Access
Microphone	Voice mode (STT)	System Settings → Privacy & Security → Microphone
Speech Recognition	Voice mode (on-device transcription)	System Settings → Privacy & Security → Speech Recognition

Important: Without Accessibility and Screen Recording permissions, desktop automation features (YukkaiPilot) will not work — clicks and typing will silently fail at the OS level.

Network Requirements

Internet connection required for:
Cloud model inference (OpenRouter, Fusion, Frontier, Free).
Web browsing and search.
Downloading local models from Hugging Face.
Stripe payment processing for Wallet top-ups.
YukkaiConnect integrations (GitHub, Slack, Notion, etc.).
Offline mode: Once a local MLX model is downloaded, inference is fully offline. Media generation and web browsing still require internet.

Storage Requirements for Local Models

Local MLX models are downloaded to ~/.cache/huggingface/hub on macOS. Approximate sizes:

Model	Download Size	RAM Required
Llama 3.2 3B	~2 GB	~4 GB
Qwen3 4B	~2.5 GB	~4 GB
Qwen3-Coder-30B	~18 GB	~20 GB
Gemma 4 27B 8-bit	~28 GB	~30 GB

Tip: The model picker shows a warning if a model is too large for your Mac's RAM before you commit to a multi-GB download.

iOS Limitations

The iOS version of Yukkai has a reduced feature set:

❌ No desktop automation — iOS doesn't allow Accessibility-style control of other apps.
❌ No bash execution — replaced by a sandboxed JavaScript runtime.
❌ No IDE integration — cannot open files in Xcode/VS Code.
⚠️ Limited local models — fewer models fit in iPhone RAM.
✅ Chat, swarm, RAG, skills, media generation — all work as on macOS.

Checking Your System

To verify your Mac meets the requirements:

# Check macOS version
sw_vers

# Check chip and RAM
system_profiler SPHardwareDataType | grep -E "Chip|Memory"

# Check free disk space
df -h /

Next Steps

First Steps — What to do after installing Yukkai.
Yukkai Local (MLX) — Running local models on your Mac.