Nodeau
Sign in Get Nodeau

Get Nodeau

Your first model, about ten minutes from now.

Pick your platform, paste a few commands, and Nodeau does the rest. It checks the machine, sets up what's needed, and starts a model with an OpenAI-compatible endpoint you can call straight away.

Free on one machine · No account needed

Linux with an NVIDIA GPU

  • Ubuntu 24.04 LTS on x86_64
  • An NVIDIA GPU with its driver working, so nvidia-smi prints your card
  • About 20 GB of free disk
  • Your password for the setup step
bash
curl -fsSL https://get.nodeau.ai/install.sh | bash
nodeau install
nodeau quickstart

nodeau install shows you its plan and every command that needs sudo, then asks. Want the plan without any changes? Run nodeau install --dry-run. If your shell needs it, the first command also prints the one line that puts nodeau on your PATH.

Linux install in detail

An Apple Silicon Mac

  • Any M-series Mac
  • Room for the models you want, starting with a 2.6 GB starter
  • No drivers to install and no password needed
  • Runs standalone, as your own private AI endpoint
zsh
curl -fsSL https://get.nodeau.ai/install.sh | bash
nodeau install
nodeau run qwen3.5-4b-q4km

Nodeau runs chat, embedding, reranking and vision models on the Mac's own GPU through Metal. Use the command above rather than a browser download. If the first command prints a PATH line, paste it, then carry on.

Mac install in detail

Like to read the installer before you run it?

Good habit! It's short, and it only downloads Nodeau, checks its checksum and puts it in ~/.local/bin. It runs as you, with no sudo.

bash
curl -fsSL https://get.nodeau.ai/install.sh -o install-nodeau.sh
less install-nodeau.sh
bash install-nodeau.sh

What happens next

Nodeau takes it from here.

It checks the machine

Operating system, GPU, driver, disks and anything Nodeau already finds installed.

It offers you a model

The starter is Qwen3.5-4B, about a 2.6 GB download under the Apache 2.0 licence. Nodeau shows the size and licence and asks first.

It checks the fit

Nodeau compares what the model needs with what your GPU can really give, before anything loads.

Your endpoint comes online

A local, OpenAI-compatible API at http://127.0.0.1:8080/v1, protected by a key made on your machine.

Say hello

Then call it like any OpenAI API.

Nodeau prints a working command when it finishes. It looks like this:

bash
export NODEAU_API_KEY="$(nodeau auth show --quiet)"

curl http://127.0.0.1:8080/v1/chat/completions \
  -H "Authorization: Bearer $NODEAU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.5-4b-q4km",
    "messages": [{"role": "user", "content": "Say hello from my local GPU."}],
    "max_tokens": 2048
  }'

Text comes back and that's your own GPU talking! The first reply takes a moment while the model warms up. For streaming, add "stream": true and use curl -N.

Use it from Python
python
import os
from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:8080/v1",
    api_key=os.environ["NODEAU_API_KEY"],
)

reply = client.chat.completions.create(
    model="qwen3.5-4b-q4km",
    messages=[{"role": "user", "content": "Say hello from my local GPU."}],
    max_tokens=2048,
)
print(reply.choices[0].message.content)

Afterwards

The commands you'll reach for.

nodeau ps

What's running on your GPUs, where, and its endpoint.

nodeau model list

What you can run, and what's already downloaded.

nodeau run <model>

Start another model, or bring a stopped one back.

nodeau stop qwen-local

Stop the endpoint. Add --workload to free the GPU too. Your downloads stay.

nodeau doctor

A read-only health check of the machine and the installation.

nodeau dashboard

Your models, services and GPU in a browser, served by your own machine.

nodeau auth show

The key your apps use for the local API. Using the API.

nodeau model import <file>

Bring a GGUF of your own. Nodeau tests what it can really do on your card. On Linux. Bring your own model.

nodeau fleet upgrade plan

Preview a new release across your machines, then apply rolls it out one at a time. Upgrading a fleet.

Every command that prints a table also takes --json, and the exit code tells a script exactly what kind of answer it got. Every command, with its flags.

More than one machine

Add another machine whenever you like.

Install Nodeau on the first Linux machine as above. Then run one command there and one on the new machine, and the two become a fleet. Several machines are part of Home Pro and Business.

Add and run machines

bash
# on the machine you already have
nodeau fleet invite

# on the new machine, within 15 minutes
curl -fsSL https://get.nodeau.ai/install.sh | bash
nodeau join <code>

If you get stuck

Send us one file and we'll take a look.

A report from a machine that's new to us is the most useful thing we get. Rather than spending an evening patching things by hand, send it our way and we'll fix Nodeau.

When the CLI runs

bash
nodeau support bundle

It writes one .tar.gz into the current folder, scrubbed of keys and credentials and free of prompts and model output. It stays on your disk until you attach it, so feel free to look inside first.

When setup stopped early

path
~/.local/state/nodeau/install.log

Everything nodeau install prints is saved there. Send the file, or paste whatever the terminal showed. Either helps.

Troubleshooting is organised by what you're seeing, and founders@nodeau.ai or the contact form reaches us directly.