GPU Endpoints

Endpoints are under /v1/resource/gpu and require authentication. GPU features must be enabled on your account.

Base URL: https://api.gpcn.com/v1

See the API Reference → for examples in TypeScript, Python, Go, and C#.

Get GPU Inventory

GET /resource/gpu/inventory

Returns real-time GPU availability by series and data center.

Query Parameters

Parameter	Type	Description
`code`	string	Filter by series code
`datacenterId`	UUID	Filter by data center
`count`	integer	Filter by GPU count (1, 2, 4, 8)

Response

{
  "success": true,
  "data": [
    {
      "seriesId": 3,
      "seriesName": "NVIDIA A100",
      "seriesCode": "a100",
      "availability": [
        {
          "datacenterId": "dc-uuid",
          "datacenterName": "US East 1",
          "gpuCounts": {
            "1": { "available": true },
            "2": { "available": true },
            "4": { "available": false },
            "8": { "available": false }
          }
        }
      ]
    }
  ]
}

Inventory responses are not cached. Each request reflects real-time availability.

curl https://api.gpcn.com/v1/resource/gpu/inventory \
  -H "X-API-Key: gpcn_your_api_key_here"

GPU inventory changes in real time. Between checking availability and placing your order, another customer may claim the last available GPU. Always handle 409 Conflict responses gracefully.

List GPU VMs

GET /resource/gpu

Returns a paginated list of GPU virtual machines.

Query Parameters

Parameter	Type	Default	Description
`page`	integer	`1`	Page number
`limit`	integer	`10`	Items per page (max 100)
`sort`	string	`createdAt:desc`	Sort field and direction
`search`	string	—	Search by name
`filter[teamId]`	integer	—	Filter by resource group

Response

Paginated array of GPU VM list items with standard pagination meta.

Create GPU VM

POST /resource/gpu

Creates a GPU VM. Returns immediately with job tracking info.

Request Body

{
  "name": "ml-training-node",
  "datacenterId": "dc-uuid",
  "seriesId": 3,
  "gpuCount": 4,
  "teamId": 5
}

Field	Type	Required	Description
`name`	string	Yes	VM name
`datacenterId`	UUID	Yes	Target data center
`seriesId`	integer	Yes	GPU series ID
`gpuCount`	integer	Yes	Number of GPUs (1, 2, 4, 8)
`teamId`	integer	No	Assign to a resource group
`imageName`	string	No	Specific image name
`imageId`	integer	No	Specific image ID

curl -X POST https://api.gpcn.com/v1/resource/gpu \
  -H "X-API-Key: gpcn_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "ml-training-01",
    "datacenterId": "dc-uuid",
    "seriesId": 1,
    "gpuCount": 2,
    "teamId": 5
  }'

Response (202 Accepted)

Job tracking object. Poll for completion.

Error: 409 Conflict

Returned when the requested GPU configuration is no longer available. Check inventory and retry.

Get GPU VM Details

GET /resource/gpu/{id}

Returns detailed GPU VM information including real-time status from the provider.

Response

{
  "success": true,
  "data": {
    "id": "gpu-vm-uuid",
    "name": "ml-training-node",
    "status": "Running",
    "publicIp": "203.0.113.50",
    "jupyterLabUrl": "https://203.0.113.50:8888",
    "series": { "id": 3, "name": "NVIDIA A100", "code": "a100" },
    "gpuCount": 4,
    "datacenter": { "id": "dc-uuid", "name": "US East 1" },
    "teamId": 5,
    "createdAt": "2026-02-21T10:00:00Z"
  }
}

Update GPU VM

PUT /resource/gpu/{id}

Update name or resource group assignment.

Request Body

{ "name": "new-name", "teamId": 5 }

Response (200 OK)

Delete GPU VM

DELETE /resource/gpu/{id}

Response (202 Accepted)

Job tracking object.

Start GPU VM

POST /resource/gpu/{id}/start

Response (200 OK)

May return 501 Not Implemented if the provider doesn't support this operation.

Stop GPU VM

POST /resource/gpu/{id}/stop

Response (200 OK)

Reset GPU VM

POST /resource/gpu/{id}/reset

Response (200 OK)

Get SSH Key

GET /resource/gpu/{id}/ssh-key

Download the SSH private key generated during VM creation.

Response

{
  "success": true,
  "data": {
    "privateKey": "-----BEGIN OPENSSH PRIVATE KEY-----\n..."
  }
}

Bulk Start

POST /resource/gpu/bulk/start

Request Body

{ "ids": ["gpu-uuid-1", "gpu-uuid-2"] }

1–50 IDs. Returns per-ID results with partial failure support.

Bulk Stop

POST /resource/gpu/bulk/stop

Same format as bulk start.

Bulk Reset

POST /resource/gpu/bulk/reset

Same format as bulk start.

Bulk Delete

POST /resource/gpu/bulk/delete

Request Body

{ "ids": ["gpu-uuid-1", "gpu-uuid-2"] }

1–50 IDs. Returns array of job tracking objects.

Workflow: Check Inventory and Create a GPU VM

GPU availability is limited and changes in real time. Here's the recommended pattern — check inventory first, then create:

# 1. Check what's available
curl -s https://api.gpcn.com/v1/resource/gpu/inventory \
  -H "X-API-Key: gpcn_your_api_key_here" | jq '.data'

# 2. Create the GPU VM (use series and datacenter from inventory)
RESPONSE=$(curl -s -X POST https://api.gpcn.com/v1/resource/gpu \
  -H "X-API-Key: gpcn_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{"name": "ml-training-01", "datacenterId": "dc-uuid", "seriesId": 1, "gpuCount": 2}')

# 3. Poll the job (same pattern as standard VMs)
JOB_ID=$(echo $RESPONSE | jq -r '.data.jobId')
# See Async Operations for the full polling pattern

For the full polling pattern, see Async Operations.

Next Steps

Create a GPU VM in the Portal → — step-by-step portal walkthrough
Manage GPU VMs → — start, stop, SSH access from the dashboard
Virtual Machine Endpoints → — standard (non-GPU) virtual machines
Async Operations — job polling patterns

Virtual Machine Endpoints

Network Endpoints