MindRelay is a macOS Menu Bar and CLI application that serves as an OpenAI-compatible API relay for Apple's Foundation Models (Apple Intelligence). It allows you to use your Mac's built-in AI capabilities with any client or tool that supports the OpenAI API format.

• OpenAI Compatible: Seamlessly integrates with tools like Chatwise and other AI clients using /v1/chat/completions.
• Privacy First: All processing happens locally on your Mac using Apple Intelligence. No data leaves your device.
• Efficient Resource Management: Automatically loads and unloads models based on activity to save memory and battery.
• Customizable: Configure the server port, and bind address.
• Menu Bar Native: Minimalist design that stays out of your way.

You can get MindRelay in two ways:

brew tap pruizlezcano/tap
brew install --cask mindrelay

Download the latest DMG or ZIP directly from the Releases page.

Access the API server from your Menu Bar and customize settings for your hardware.

Using MindRelay with Chatwise to generate code and content locally.

• macOS Tahoe (26.0+)
• Apple Intelligence enabled and available on your device.

MindRelay automatically starts the API server when launched. You can check the status and copy the server URL directly from the Menu Bar.

Go to Settings from the Menu Bar to configure:

• Port: Change the port if 11343 is in use.
• Bind Address: Choose between localhost or 0.0.0.0 for local or network access.
• Model Unload Timeout: Set how long the model stays in memory after the last request.

• GET /health: Check server status.
• GET /v1/models: List available models (currently returns a static list with apple-foundation-model).
• GET /v1/models/{model_id}: Get details about a specific model.
• POST /v1/chat/completions: Main endpoint for generating chat completions using Apple Intelligence.

curl http://localhost:11343/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "apple-foundation-model",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ]
  }'

MindRelay includes an integrated CLI for direct interactions and automation.

• Auto with Homebrew: If you installed MindRelay via Homebrew, the mindrelay CLI tool is automatically linked and ready to use in your terminal.
• Manual Installation: If you downloaded the app manually, you can install the CLI tool directly from the app's Settings.

Prompt Send a direct prompt to the model and get the entire response at once:

mindrelay "What is the capital of France?"

Stream Prompts Stream the response continuously to stdout as it's generated:

mindrelay --stream "Write a short poem about space."

Pipe Pipe stdin directly into the AI for processing or summarizing:

echo "Hello" | mindrelay

Server Start the OpenAI-compatible HTTP server directly from the terminal (useful for headless environments or automation):

mindrelay --serve

Note: If no configuration options are provided (like --port, --addr, or --model-unload-timeout), the CLI will automatically use the defaults configured in the MindRelay app settings.

If you wish to build and run MindRelay from source for development purposes:

1. Clone the repository: git clone https://github.com/pruizlezcano/MindRelay.git
2. Open MindRelay.xcodeproj in Xcode.
3. Select your development team in Signing & Capabilities.
4. Press Cmd + R to Build and Run.

All data processing happens locally on your Mac using Apple Intelligence. MindRelay does not send any data to external servers. No user data is collected or stored by MindRelay.

• Server Not Starting: Ensure no other application is using port 11343. Change the port in Settings if needed.
• Model Not Loading: Check that Apple Intelligence is enabled and available on your Mac.
• API Errors: Use the /health endpoint to check server status and review logs for detailed error messages.
• You can see all the logs from the Settings -> About section.

For support or feature suggestions, visit the GitHub Issues page.