Skip to main content

Quickstart

Follow these steps to get a node online, configure the dashboard, and send your first request.

note

The node agent always uses the embedded llama.cpp backend. If you build it from source, initialize third_party/llama.cpp and build the support library first.

note

Cloudless AI currently supports macOS with Apple silicon. Support for Linux and Windows is coming soon.

1. Install the node agent​

Install Cloudless AI globally on any GPU or NPU machine:

npm install -g @cloudless-ai/node-agent

2. Start the node agent​

Run the following command to start the node agent:

cloudless-node start

3. Log in to your Cloudless AI account​

Follow the prompts to log in or create a Cloudless AI account. The agent will register with your account automatically.

4. Open the dashboard​

Go to the Cloudless AI dashboard and check Overview.

  • Confirm the node appears online.
  • Confirm the router can see its hardware and loaded models.

5. Add a cloud fallback​

Open Settings > Cloud Fallback and add a cloud provider as your fallback:

  • OpenAI
  • Anthropic
  • Microsoft Foundry (OpenAI SDK or Inference SDK)

For each provider:

  • Give it a display name.
  • Add the provider base URL.
  • Set the API key.
  • Choose the models or deployments it can serve.
  • Click Test and then Save.

6. Create a routing rule​

Open Routing and create a rule for the model names your app will request.

  • Match the incoming model name.
  • Choose which node models to try first.
  • Choose which cloud fallback models are allowed.
  • Save the rule.

7. Send a request from your app​

Point your app's OpenAI or Anthropic SDK at the router URL and send a normal request.

See Connect your app for examples.

Next steps​