Hot to connect headless to nucleus

We´ve setup a dedicated GPU-server under linux. This system is headless, so we cannot open (for example) Create to establish the connection to our Nucleus.

Can anyone explain how to add a Nucleus connection in such cases?

Thanks for help


hello @c.bickmeier -
Can you give me more details on how you setup the server to run Create. Are you using containers? starting up Create from command line in headless mode? how are you getting the GUI.

At first: We do not use Create that was a little bit misleading… we have an own kit app, that is only a renderer with an extension we have written to control that renderer.

The GUI comes per websocket streaming (at the moment, we will switch to webrtc later).

The problem is, that the server does not have any display connected. As far as I understood, that prevents us from opening the streaming-kit-app on the GUI-console (Q6 from here Linux Troubleshooting — Omniverse Kit documentation (

Our render-app has to load an usd-scene from our nucleus, so we have to establish a connection to that nucleus. But on the desktop we cannot open the render app (problem from above). And if we would try to connect through the streamed gui, we cannot “switch” to a browser to enter the credentials.

So I`m looking for a way to authenticate a headless kit process against a nucleus server. Is it (for example) possible to copy the token form another app that already has a successful login on the nucleus to the config files of my render-app (to prevent the roundtrip through the browser, when streaming the UI)?

Or are there any environment variables I can set for user and password for a specific nucleus server?

How does kit farm for example handle this?

Thanks for your help


I just remebered myself the the connector sample works with environment variables (OMNI_USER and OMNI_PASS), so I tried to set those Variables for the render processe… now the messages have changed, but I still cannot connect…

What dos “Connecting to … as username using NEW authentication method”?

Sorry… but I just found this…

API Tokens — Omniverse Nucleus documentation (

Will try that later (at the moment I cannot access the nucleus)

Why du I see the api token menu on my local nucleus…
…but not on our enterprise nucleus?

hello @c.bickmeier -
Enterprise Nucleus versions 2022.x support this. I am guessing that you are running 2021.3.2 or previous. Can you tell me the version to verify?

Yes… our nucleus is too old, but our test license is outdated… try to contact my nvidia-contact to extend the test license…

Thank you