Hanzo

Machines API

Launch, inspect, and destroy machines and GPUs through api.hanzo.ai/v1.

A machine is a compute unit — a Visor-provisioned VM, or a bring-your-own worker that dialed in with hanzo link. Both appear in the same list, distinguished by provider.

All calls take an IAM bearer — see Authentication.

Endpoints

MethodPathDescription
GET/v1/machinesList every machine the org can see
POST/v1/machinesLaunch a machine
GET/v1/machines/{id}Machine detail
DELETE/v1/machines/{id}Destroy a machine
GET/v1/gpusEvery GPU across the org's compute
GET/v1/gpus/alertsGPU health alerts
GET/v1/fleetThe unified board — every compute source with its latest utilization
GET/v1/fleet/workersBYO workers only
GET/v1/fleet/jobsThe org's GPU job queue
POST/v1/fleet/jobs/{id}/cancelCancel a queued or running job
GET · POST/v1/fleet/samplesRead back / self-report utilization samples

List machines

curl -fsS https://api.hanzo.ai/v1/machines \
  -H "Authorization: Bearer $TOKEN"
{
  "machines": [
    {
      "id": "dbc",
      "name": "dbc",
      "region": "on-prem",
      "type": "byo-gpu",
      "status": "offline",
      "provider": "byo"
    }
  ]
}

List GPUs

curl -fsS https://api.hanzo.ai/v1/gpus \
  -H "Authorization: Bearer $TOKEN"
{
  "gpus": [
    {
      "id": "dbc#0",
      "name": "dbc",
      "model": "Apple M4 Max (Metal)",
      "region": "on-prem",
      "status": "offline"
    }
  ]
}

Destroy a machine

curl -fsS -X DELETE "https://api.hanzo.ai/v1/machines/$MACHINE_ID" \
  -H "Authorization: Bearer $TOKEN"

Destroying a machine is permanent. Detach or snapshot any volume you still need first.

How is this guide?

Last updated on

On this page