Dedicated Clusters

An organization on an enterprise plan can have one or more dedicated clusters: Spice-managed, single-tenant infrastructure where an organization's projects run only alongside other projects from the same organization — never on shared infrastructure. Each cluster has its own cluster_name and isolated network.

Dedicated clusters are provisioned by Spice.ai and requested through support. Once a cluster is provisioned and registered to an organization, it is available to the Management API and in the Portal's project-creation picker.

Listing clusters

GET /v1/clusters returns the dedicated clusters registered to the organization bound to the access token. It requires the apps:read scope.

curl -H "Authorization: Bearer <token>" \
  https://api.spice.ai/v1/clusters

{
  "clusters": [
    {
      "cluster_name": "acme-prod-sandbox",
      "region": "us-west-2",
      "cloud_provider": "aws",
      "endpoint": "https://private-acme-prod-sandbox-us-west-2-prod-data.spiceai.io",
      "created_at": "2026-06-11T00:00:00Z",
      "updated_at": "2026-06-11T00:00:00Z"
    }
  ]
}

• cluster_name — the cluster's identifier, used when creating or reassigning projects.
• endpoint — the cluster's data-plane endpoint (an https:// URL); projects running on the cluster are reached at this URL. More than one cluster in the list can report the same endpoint, so match a cluster on cluster_name rather than on endpoint.

Creating a project on a dedicated cluster

A create request specifies cluster_name instead of region, set to a cluster_name returned by GET /v1/clusters. Exactly one of the two is provided; the project's region is derived from the cluster. The request requires the apps:write scope.

curl -X POST https://api.spice.ai/v1/projects \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-project",
    "cluster_name": "acme-prod-sandbox",
    "description": "A project on a dedicated cluster"
  }'

The response includes the resolved assignment and the cluster's endpoint:

{
  "id": 123,
  "name": "my-project",
  "cluster_name": "acme-prod-sandbox",
  "endpoint": "https://private-acme-prod-sandbox-us-west-2-prod-data.spiceai.io",
  "...": "..."
}

A project created without cluster_name deploys to the shared regional infrastructure as usual; cluster_name: null is equivalent to omitting it.

Errors

Moving an existing project to a dedicated cluster

PUT /v1/projects/{projectId} with cluster_name reassigns the project. Subsequent deployments land on the cluster, and the project's endpoint changes to the cluster's host.

curl -X PUT https://api.spice.ai/v1/projects/123 \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{"cluster_name": "acme-prod-sandbox"}'

Querying projects on a dedicated cluster

The project and GET /v1/clusters responses return the cluster's endpoint — the URL clients connect to.

It serves the same APIs (SQL, search, and LLM over HTTP, plus Apache Arrow Flight), and authentication is unchanged — the project's API key or platform credentials work exactly as on shared infrastructure. For Apache Arrow Flight, the endpoint's host is used with -data replaced by -flight, over grpc+tls://<host>:443.

With the SDKs, the endpoint replaces the data.spiceai.io / flight.spiceai.io defaults. For example, with the Python SDK over Flight:

from spicepy import Client

client = Client(
    api_key="<project-api-key>",
    url="grpc+tls://private-acme-prod-sandbox-us-west-2-prod-flight.spiceai.io:443",
)

Everything else — deployments, secrets, API keys, spicepod configuration — works identically to projects on shared infrastructure. The Management APIs reference documents the full endpoint set.