Skip to main content
This guide takes you from a new account to a live model responding to an API call.
The dashboard is at valkyrieapp.azumo.com and the API base is https://valkyrie-back.azumo.com. If your organization runs a separate environment, use its hosts instead.

1. Create an account and fund your wallet

  1. Sign in at valkyrieapp.azumo.com.
  2. You land on Home. Open Wallet and check (or top up) your balance. Your balance is also shown in the Balance chip in the top bar. Valkyrie is prepaid and billed per use, your balance is what your deployments and jobs draw against. See Wallet & billing.

2. Create an API key

Machine access uses an API key sent in the X-API-Key header.
  1. Open API keys from the sidebar, then the Deployment Keys tab.
  2. Click Create API key and copy it, you won’t be able to see it again. Choose Account-wide if you haven’t deployed anything yet; to scope a key to one endpoint, create it once that endpoint is starting or running.
Treat API keys like passwords. Store them in a secret manager or environment variable, never in source control.

3. Deploy a model

You can deploy any Hugging Face model directly, or deploy a model you’ve fine-tuned on Valkyrie.
  1. Open Endpoints and click Deploy a base model. To deploy a fine-tuned model instead, open it on Models, select a completed version, and click Deploy.
  2. Choose a preset (for example, a Qwen chat model) or enter any Hugging Face model path.
  3. Launch the deployment. Valkyrie provisions a GPU on demand, starts the serving engine, and health-checks it before marking it Running.
When the deployment is ready it exposes an OpenAI-compatible endpoint. Copy its URL from the endpoint’s page, and try it in the Playground before writing code. See Deploy a model for the full walkthrough.

4. Call your model

Valkyrie speaks the OpenAI Chat Completions API. Point any OpenAI client at your deployment’s base URL and use your API key.
Prefer a stable, readable URL? Attach an alias to your deployment and call /{account_slug}/{alias}/v1/chat/completions instead of the raw deployment slug. See Aliases.

Next steps

Fine-tune a model

Train an open model on your own dataset.

Fine-tune with the AI assistant

Attach a spreadsheet, say what to predict, and confirm.

Connect to Claude Code

Use your Valkyrie model from your coding agent.

Automate with MCP

Provision and operate models from an AI agent.

Understand billing

How the prepaid wallet and per-use pricing work.