Documentation Index
Fetch the complete documentation index at: https://mintlify.com/jundot/omlx/llms.txt
Use this file to discover all available pages before exploring further.
oMLX runs on macOS 15.0+ (Sequoia) with Apple Silicon (M1/M2/M3/M4) and Python 3.10+. Choose the installation method that fits your workflow: the macOS app is the fastest way to get started, Homebrew gives you CLI access and background service management, and source install is best if you want to modify oMLX itself.
macOS App
Homebrew
From Source
Download the .dmg from GitHub Releases, open it, and drag oMLX to your Applications folder.Once installed, launch oMLX from Applications. The Welcome screen walks you through three steps: set your model directory, start the server, and download your first model.The macOS app includes in-app auto-update — future upgrades are a single click inside the app.The macOS app does not install the omlx CLI command. If you need terminal access (for example, to use omlx serve or omlx launch), install via Homebrew or from source instead.
Homebrew gives you the omlx CLI and lets you run oMLX as a managed background service that auto-restarts on crash.brew tap jundot/omlx https://github.com/jundot/omlx
brew install omlx
Running as a background service
After installing, you can start oMLX as a system service:brew services start omlx # Start (auto-restarts on crash)
brew services stop omlx # Stop
brew services restart omlx # Restart
brew services info omlx # Check status and PID
The service runs omlx serve with zero-config defaults: model directory ~/.omlx/models, port 8000. To customize, set environment variables before starting the service, or run omlx serve --model-dir /your/path once to persist settings to ~/.omlx/settings.json.Environment variables
| Variable | Description | Default |
|---|
OMLX_MODEL_DIR | Directory containing your model subdirectories | ~/.omlx/models |
OMLX_PORT | Port the server listens on | 8000 |
Log locations
oMLX writes logs to two locations:
- Service log:
$(brew --prefix)/var/log/omlx.log — stdout/stderr from the service process
- Server log:
~/.omlx/logs/server.log — structured application log with request details
MCP support (optional)
To enable Model Context Protocol tool integration, install the mcp package into oMLX’s isolated environment:/opt/homebrew/opt/omlx/libexec/bin/pip install mcp
Upgrading
brew update && brew upgrade omlx
Install from source if you want to modify oMLX or contribute to development.git clone https://github.com/jundot/omlx.git
cd omlx
pip install -e .
# With MCP (Model Context Protocol) support
pip install -e ".[mcp]"
# With development tools (pytest, black, ruff, mypy)
pip install -e ".[dev]"
# With grammar-constrained decoding (requires ~2 GB for torch)
pip install -e ".[grammar]"
The source install supports Python 3.10, 3.11, 3.12, and 3.13. For the macOS app bundle, Python 3.11+ is required (see packaging/README.md).
After installing, run the server directly:omlx serve --model-dir ~/models
Verifying the installation
After installing via Homebrew or from source, confirm the CLI is available:
Then start the server and check the health endpoint:
omlx serve --model-dir ~/models
curl http://localhost:8000/health
You should receive a 200 OK response. The admin dashboard is available at http://localhost:8000/admin.