Skip to main content

How it Works

CaseDesk deploys open-source AI models and gives you an OpenAI-compatible API endpoint. You can use Managed CaseDesk (CaseDesk runs the cluster) or connect your own GPU VM or on-prem server.

Managed CaseDesk​

CaseDesk provisions a dedicated cluster in your chosen region and manages it for you. Flat monthly subscription — no GPU-hour billing.

  1. Start the wizard — go to getcasedesk.com and click Get started
  2. Choose a plan — Starter (1–8B models), Team (9–20B, up to 20 concurrent users), or Advanced (21–70B, up to 50 concurrent users)
  3. Choose a region — UK (AWS eu-west-2, DSPT compliant), EU (Azure westeurope), or US (GCP us-east1)
  4. Choose a model — pick from the catalogue (Llama, DeepSeek, Qwen, Phi, and more)
  5. Use the endpoint — every deployment gets a unique OpenAI-compatible URL and a production API key (cd_live_...)

Managed deployments scale to zero when idle and wake automatically on the next request. See Autoscaling for cold-start times.

GPU VM (single cloud server)​

Connect an existing cloud server. CaseDesk installs the runtime and proxies requests through your dashboard URL.

  1. Connect your provider — sign in with Google/Microsoft/DigitalOcean OAuth, or paste an API key for AWS/Hetzner/Lambda Labs. CaseDesk lists your running VMs automatically. You can also enter an IP address manually.
  2. Select a VM — pick from the list; the IP address and provider are pre-filled
  3. Upload your SSH key — CaseDesk encrypts it with AES-256-GCM and uses it once to bootstrap
  4. CaseDesk bootstraps — installs Docker, runs Ollama, pulls the selected model
  5. Use the endpoint — proxied through your CaseDesk dashboard URL

On-prem server (outbound agent, no inbound ports)​

For servers inside your network where inbound SSH is not possible (air-gapped environments, HSCN-connected NHS trusts, corporate firewalls):

  1. Generate an install token — CaseDesk creates a one-time token valid for 24 hours
  2. Run the one-liner on your server — the agent installs itself and connects outbound over HTTPS
  3. Agent registers — CaseDesk receives the connection and marks the server as live
  4. Deploy a model — same catalogue, same one-click deploy

The agent never requires an inbound port. All traffic is initiated by the server to CaseDesk.

Your data never leaves your infrastructure.