Prerequisites

Before you add a node, make sure your machine meets these requirements. You'll run a small agent (a single Docker command) that connects your machine to the network — outbound only, so no port forwarding or public IP is required.

Hardware

You'll need an NVIDIA GPU with at least 8 GB of VRAM, 16 GB RAM, 50 GB of free SSD space, and a 50 Mbps connection. For the full minimum-vs-recommended breakdown, see Hardware Requirements.

Software

• Linux or Windows.
• Up-to-date NVIDIA drivers.
• Docker — install guide.
• The NVIDIA Container Toolkit, so containers can access the GPU — install guide.

Verify your setup

Confirm Docker can see your GPU before you continue:

docker run --rm --gpus all nvidia/cuda:12.4.0-base-ubuntu22.04 nvidia-smi

If your GPU shows up in the output, you're ready to Add a Node.

Troubleshooting

invoking the NVIDIA Container Runtime Hook directly (e.g. specifying the docker --gpus flag) is not supported, with Using requested mode 'cdi' before it. The NVIDIA Container Toolkit is set to CDI mode, which doesn't accept Docker's --gpus flag; the node stays online but its GPU verification fails. Put the toolkit back in automatic mode and restart Docker:

sudo nvidia-ctk config --in-place --set nvidia-container-runtime.mode=auto
sudo systemctl restart docker

Then run the check above again. Once it prints your GPU, the node verifies itself within a few minutes.

could not select device driver "" with capabilities: [[gpu]]. Docker doesn't know about the NVIDIA runtime yet. Install the NVIDIA Container Toolkit, then:

sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker