macOS
Apple Silicon, M1 or newer
curl --proto '=https' --tlsv1.2 -fsSL https://gethephaistos.com/install.sh | bash
Uses MLX and the Mac GPU automatically.
Personal AI node
One command detects your hardware, refreshes the signed model catalog, proposes three compatible models and prepares secure remote access. You only choose a model, then enter the pairing code in Hephaistos.
One command
Run the command on the machine that will host Jarvis. The installer uses a checksum-verified closed package; it does not download the private application source or request infrastructure credentials.
Apple Silicon, M1 or newer
curl --proto '=https' --tlsv1.2 -fsSL https://gethephaistos.com/install.sh | bash
Uses MLX and the Mac GPU automatically.
x86_64 or arm64
curl --proto '=https' --tlsv1.2 -fsSL https://gethephaistos.com/install.sh | bash
Uses Vulkan when available, with a CPU path otherwise.
64-bit x86_64
powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://gethephaistos.com/install.ps1 | iex"
Uses Vulkan when available, with a CPU path otherwise.
Automatic setup
Hephaistos reads the operating system, processor, memory and available GPU acceleration.
The signed catalog is downloaded and reduced to exactly three choices that fit this machine.
Hephaistos explains the recommended, faster and lighter options before installing anything large.
The model engine runs only on local loopback. An outbound encrypted tunnel connects it to your account without opening a router port.
In the app, open Settings, Server setup, Your server, then enter the 10-character code shown by the installer.
Hephaistos Cloud is the paid managed alternative.