Add and manage upstreams

An upstream is one LLM API that Noket can reach from this machine: its base URL, its protocol (OpenAI or Anthropic compatible), its key if it needs one, and the model ids you want to expose.

You need three things before you start:

1. Noket installed and logged in — see Install and Log in.
2. An LLM server running on this machine or on your network, with an HTTP API.
3. The base URL of that API.

Add an upstream

noket upstream add

The command opens a form. Fill in:

After the first field is filled, the form probes {base_url}/v1/models and lists what the server reports. Select the models you want; typing a model id by hand is also accepted. If the probe fails, you can still save the upstream — check the URL and try noket upstream test afterwards.

Add an upstream without prompts

Setting any of --name, --protocol, --url, --model, or --api-key-stdin switches the command to non-interactive mode:

noket upstream add \
  --name ollama \
  --protocol openai \
  --url http://127.0.0.1:11434 \
  --model llama3.2 \
  --model qwen2.5

--name, --url, and at least one --model are required. For a server that needs a key, pipe it in from your secret store:

op read "op://Private/vllm/api-key" | noket upstream add --name vllm --url http://127.0.0.1:8000 --model my-model --api-key-stdin

Any command that writes the key to standard output works — pass show, security find-generic-password -w, or the equivalent on your system.

If an upstream with that name already exists, noket upstream add updates it instead of failing.

Check it

noket upstream list
noket upstream test ollama

list shows each upstream’s name, protocol, base URL, models, and only whether a key is set. test probes the API and prints the models it discovers, without sending a generation request.

Start the tunnel

noket up

This starts noketd in the background. The command prints the process id, the node id and name, and the log path. The node id is generated on the first start and stays with this machine.

The connector now keeps a WebSocket connection open to Noket. Requests that arrive for this node are forwarded to the upstream that serves the requested model id.

Create a model in the console

The tunnel carries requests, but you still need a model in the console to call. Go to Models → New model, choose Tunnel node as the source, pick this node, and choose one of the upstream model ids you exposed. The console then gives you a model id you can call with your API key.

Next

• Run a local server — Ollama, vLLM, LM Studio, and Anthropic-compatible servers
• Manage upstreams — every flag of add, edit, list, test, and remove
• How the tunnel works