Available in VPC
Describes the Overview interface of Project Home. On the Overview tab, you can download kubeconfig to access the Project Namespace and view recent task history.

Download kubeconfig
Each project has its own Kubernetes Namespace, which provides an environment isolated from other projects in the Workspace.
Project Namespace is created in the p-{프로젝트 이름} format.
A kubeconfig file that can access the Namespace is required to create or view workloads in the Project Namespace using kubectl. The kubeconfig contains your own token, and the token's permissions follow your Project member role (Admin, Writer, Reader).
Issue kubeconfig
To issue kubeconfig:
- At the top of the Overview tab in Project Home, click Download Kubeconfig.
- In the Kubeconfig download popup, enter the token duration (days) in Token Validity Period (days).
- The default value is 90 days; enter it as a whole number of days.
- The expiration time is displayed below the input value.
- Click Download.
{프로젝트 이름}-kubeconfig.yamlThe file will be downloaded.
- Unlike API Keys, issued kubeconfig tokens cannot be revoked before they expire. Instead of specifying an unnecessarily long validity period, if the file is leaked, ask the Project Admin to remove the user from the Project members. If removed from members, you cannot access the Project Namespace even if the token has not expired.
- Only you should use your kubeconfig. Do not share the file with other users or external programs, and issue an API Key for external programs.
- The maximum validity period may vary depending on the platform's operation settings. The default upper limit is 365 days, and requests exceeding this limit are denied.
- Once the validity period passes,
kubectlrequests are denied with a 401 error. If denied, download kubeconfig again from the Overview page and replace the existing file with the new one.
Check expiration time
The expiration time is recorded in a comment at the top of the kubeconfig file.
head {project name}-kubeconfig.yaml
# Expires : 2026-12-09 14:30 KST (valid for 90 days)
# Check expiry : head kubeconfig.yaml (this comment)
Accessing ML expert Platform with kubectl
You can use ML expert Platform via the CLI for Kubernetes cluster control kubectl.
kubectlThe version must be within 1 minor version of the cluster's Kubernetes version (kubectl Version Skew Policy).- Check the cluster's Kubernetes version with the
kubectl versioncommand after configuring kubeconfig.
Move the downloaded kubeconfig file to an arbitrary directory (e.g., $HOME/.kube) and specify the file path in the KUBECONFIG environment variable.
mkdir -p ~/.kube
mv {download path}/{project name}-kubeconfig.yaml ~/.kube
export KUBECONFIG=$HOME/.kube/{project name}-kubeconfig.yaml
You can use kubectl to verify that you can access the project as follows:
$ kubectl -n p-{project name} get pods
NAME READY STATUS RESTARTS AGE
ml-pipeline-ui-artifact-7496857fcc-p24rw 2/2 Running 0 5h3m
ml-pipeline-visualizationserver-5c96659f4b-d6stz 2/2 Running 0 5h3m
If you use multiple projects, specify all the kubeconfig files for each project in KUBECONFIG by connecting them with a colon (:), and then switch with the kubectl config use-context command.
For more information about how to access Kubernetes using kubeconfig, see the official documentation.
Scope of access by role
The tasks you can perform with kubeconfig vary depending on the Project member role. Any changes to the role are reflected immediately without having to receive a new kubeconfig.
| Role | Tasks that can be performed |
|---|---|
| Admin | Manage all Writer tasks and API keys. |
| Writer | Create, edit, and delete workloads (Pod, Job, PyTorchJob, Notebook, InferenceService, etc.) and ConfigMaps, Secrets, and PVCs in the project namespace. |
| Reader | View resources in the project namespace. |
If a permission error occurs, check the role in Members or contact the Project Admin.