Connect iOS devices to local AI services running on your Mac with zero configuration.

Elmer is a minimalist system that bridges iOS and macOS, allowing your iPhone or iPad to securely access AI services running locally on your Mac. No exposed ports, no complex networking—just scan a QR code and start using your local AI infrastructure from anywhere.

• 🔐 Zero-config security - Uses iCloud private database as encrypted relay
• 📱 Native iOS interface - Chat with LLMs, generate images, run custom tools
• 🤖 Auto-discovery - Detects ComfyUI, Ollama, LM Studio automatically
• 🛠️ Custom tools - Define your own command-line tools and scripts
• ☁️ No external services - Everything routes through your iCloud account
• 📡 MCP integration - Connect to Model Context Protocol servers

┌─────────────────┐    ┌─────────────────────┐    ┌─────────────────┐
│                 │    │                     │    │                 │
│   iPhone/iPad   │◄──►│   iCloud Private    │◄──►│   Mac (Server)  │
│   (Client)      │    │   Database (Relay)  │    │                 │
│                 │    │                     │    │                 │
└─────────────────┘    └─────────────────────┘    └─────────────────┘

1. Mac app runs as menu bar service, detects local AI services
2. QR code contains connection info for easy device pairing
3. iOS app sends requests via CloudKit to Mac
4. Mac processes requests locally and returns responses
5. Push notifications deliver responses back to iOS instantly

• macOS device with AI services (Ollama, ComfyUI, LM Studio, etc.)
• iOS device signed into same iCloud account
• Xcode for building apps

cd elmerMac/elmer
xcodebuild -scheme elmer -configuration Release build
open build/Release/elmer.app

cd elmerMobile/elmer
xcodebuild -scheme elmer -sdk iphonesimulator build
# Or open in Xcode and run on device

1. Mac: Click menu bar icon → "Generate QR Code"
2. iOS: Open Elmer app → Scan QR code
3. Done! Your devices are now connected

• Tap any detected language model service
• Start chatting immediately with native iOS interface
• Supports text, images, and multimodal conversations

• Access ComfyUI workflows through dedicated interface
• Generate and save images directly to Photos app

• Use predefined tools or create your own
• Execute terminal commands, scripts, and automations
• See USER_TOOLS_GUIDE.md for tool creation

Create your own tools by adding JSON files to ~/.elmer/tools/:

{
  "name": "System Info",
  "description": "Get macOS system information",
  "command": ["system_profiler", "SPSoftwareDataType"],
  "category": "system"
}

See USER_TOOLS_GUIDE.md for complete documentation.

Auto-detected:

• Ollama (http://localhost:11434)
• ComfyUI (http://localhost:8188)
• LM Studio (http://localhost:1234)
• Text Generation WebUI (http://localhost:7860)

Manual configuration:

• Any OpenAI-compatible API
• Custom web services
• MCP servers

elmer/
├── elmerMac/           # macOS menu bar application
│   └── elmer/
│       ├── ServiceManager.swift      # AI service detection
│       ├── CloudKitRelayManager.swift # Request processing
│       ├── UserToolManager.swift     # Custom tools
│       └── MCPServerManager.swift    # MCP integration
├── elmerMobile/        # iOS client application
│   └── elmer/
│       ├── UnifiedControlPanelView.swift # Main interface
│       ├── CloudKitRelayClient.swift     # Request sending
│       └── RelayConnectionManager.swift  # Connection handling
├── example-tools/      # Sample tool definitions
└── docs/              # Documentation

• End-to-end encryption via iCloud's private database
• No external servers - all communication through Apple's infrastructure
• Local processing - AI requests never leave your devices
• Automatic cleanup - Old requests deleted automatically
• iCloud sync - Works across all your signed-in devices

Both apps are standard Xcode projects:

# Mac app
cd elmerMac/elmer && xcodebuild -scheme elmer build

# iOS app
cd elmerMobile/elmer && xcodebuild -scheme elmer -sdk iphonesimulator build

1. Add detection logic to ServiceDetector.swift
2. Update service definitions in Models.swift
3. Test auto-discovery in Mac app

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Test on both Mac and iOS
4. Submit pull request

macOS:

• macOS 13.0+ (Ventura)
• iCloud account
• Xcode 15.0+ (for building)

iOS:

• iOS 16.0+
• Same iCloud account as Mac
• Camera access (for QR scanning)

Connection Issues:

• Verify both devices use same iCloud account
• Check iCloud is enabled in System Preferences/Settings
• Restart both apps if QR scanning fails

Service Detection:

• Ensure AI services are running on default ports
• Check Console.app for service detection logs
• Manually add services if auto-detection fails

Performance:

• CloudKit sync may take 1-2 seconds in some regions
• Large responses (images) may take longer to transfer
• Check network connectivity if requests timeout

• USER_TOOLS_GUIDE.md - Creating custom tools
• PROJECT_SUMMARY.md - Architecture overview
• MIGRATION_NOTES.md - Version history
• CLOUDKIT_IMPLEMENTATION.md - Technical details

MIT License - see LICENSE file for details.

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: [email protected]

Elmer: Your AI services, everywhere.