NeboAI · how to

Install the runtime

Get Nebo running on macOS, Windows or Linux. Inference works the moment you sign in, with no API key to fetch first.

TimeFive minutes
You needNothing but the installer

Install it

On macOS, through Homebrew:

brew install --cask neboloop/tap/nebo

On Windows and Linux, take the installer from the download page. All three platforms run the same runtime, and the same employees run on any of them.

Give it a model

Sign in and you are done. That is not a simplification: signing in wires the Right Fit Gateway, which fits each request to a model that suits it, and inference works immediately with no key of your own and nothing to configure.

Most tools in this category stop you here and make you go and get an API key first. You can do that if you want to, and there are good reasons to, but it is a choice rather than a step.

OptionWhat you doWhy
The gatewayNothing. Sign in.It works out of the box, and requests get fitted to a suitable model rather than all going to one.
Your own keyAdd a key for OpenAI, Anthropic, Google or DeepSeekYou pay the provider directly, on their terms, with us nowhere in the path.
OllamaRun it, pull a modelNothing leaves the machine at all, not even the model call.
Your own key wins when you add one

These are not exclusive. Add a key later and it takes priority over the gateway automatically, so you can start with zero configuration and move to your own account whenever it suits, without rebuilding anything. Ollama is picked up on its own if it is running, no key and no profile needed.

Confirm it works

Open the app and ask it something you can check. You are not testing the model, you are testing that the provider is wired up. If the answer comes back, you are done.

If nothing comes back

Check you are signed in first, since that is what wires the gateway. If you added your own key, re-paste it: an invalid key tends to fail quietly rather than loudly, and it takes priority over the gateway, so a bad key can break something that worked a minute ago. On Ollama, check the model is pulled and the service is running.

What you now have

A working runtime with nothing hired yet. Conversations, files and memory live on your own disk in SQLite. The only thing that has left the machine is the inference call, and on Ollama not even that.

Next: hire something, or connect a mailbox for it to work from.