vCluster, vCluster Platform, and vMetal work with the hardware, AI software, and operations tools you already run. This page collects every documented integration in one place, and each card leads to the guide that sets it up. To see what one partner's integrations do together, start with its section, such as NVIDIA.
Each card names the docs it leads to: vCluster, vCluster Platform, or vMetal. A Free or Enterprise tag means the integration needs a platform license at that tier. See Licensing for the plans. An Experimental tag marks a feature that can change between releases. Tags repeat what each guide states, so a missing tier tag means the guide doesn't state one.
A card means the docs cover setting up that integration. It doesn't state a formal partnership or a certification.
NVIDIA
NVIDIA integrations cover each layer of a GPU cloud. Use any of them on its own, or combine them:
- Provision GPU infrastructure with NVIDIA Infra Controller (NICo) through vMetal.
- Prepare GPU nodes and collect GPU metrics with the NVIDIA GPU Operator and DCGM.
- Create and connect Kubernetes clusters with NVKE.
- Schedule and serve workloads with the KAI scheduler, NVIDIA Run:ai, or NVIDIA Dynamo.
- Watch GPU health with the Fleet Intelligence Agent and NVSentinel.
The NVIDIA DSX OS guide shows how these pieces fit together on the platform. Each NVIDIA card also appears in the section below that matches what it does.
layersNVIDIA DSX OSvCluster PlatformSee how NICo provisioning, GPU health tooling, and the bundled NVIDIA Certified Stacks fit together in the platform.memoryNVIDIA Infra ControllervMetalEnterpriseConnect an existing NVIDIA Infra Controller (NICo) site, and provision its machines as instances and private nodes.scheduleNVIDIA KAI schedulervClusterSchedule GPU and CPU workloads in a cluster with queuing, fractional GPU allocation, and topology awareness.model_trainingNVIDIA Run:aivCluster PlatformDeploy NVIDIA Run:ai with a dedicated or central control plane for its tenants, from a Certified Stack.boltNVIDIA DynamovCluster PlatformInstall the NVIDIA Dynamo inference runtime into a cluster or control plane cluster, so it can serve models across GPU nodes.account_treeNVIDIA NVKE management planevCluster PlatformInstall the NVKE management plane into a cluster on private nodes from a signed release. NVKE is an upstream preview for evaluation.linkNVKE clustersvCluster PlatformConnect clusters that NVKE created to the platform, deploy applications into them, and configure metrics.speedNVIDIA GPU observabilityvCluster PlatformFreeDeploy the NVIDIA GPU Operator, DCGM metrics, and the Fleet Intelligence Agent to GPU clusters through Argo CD templates.health_and_safetyNVIDIA NVSentinelvCluster PlatformFreeDetect GPU node faults and respond by cordoning, draining, and rebooting nodes, deployed through an Argo CD template.trending_upGPU and inference autoscalingvClusterEnterpriseScale workloads with the Horizontal Pod Autoscaler (HPA) and KEDA, using NVIDIA DCGM GPU metrics and model-serving metrics.
Supply worker nodes, control plane machines, and standalone servers from the hardware and hypervisors you operate.
KubeVirt appears three times, depending on what the virtual machines are for. The node provider creates virtual machines that join a cluster as its worker nodes. Virtual machine management lets administrators run and operate virtual machines from the platform UI. The vCluster integration lets workloads inside a cluster create their own virtual machines.
dnsMetal3vMetalEnterpriseProvision bare metal servers with Metal3 and Ironic, as private nodes, cluster control planes, or standalone servers.memoryNVIDIA Infra ControllervMetalEnterpriseConnect an existing NVIDIA Infra Controller (NICo) site, and provision its machines as instances and private nodes.cloud_syncExternal platformvMetalEnterpriseSupply private nodes from bare metal servers that another vCluster Platform manages, without access to its Metal3 infrastructure.inventoryNetBoxvCluster PlatformEnterpriseSynchronize bare metal machine inventory from NetBox for Metal3 provisioning.developer_boardKubeVirt: VMs as nodesvMetalFreeProvision KubeVirt virtual machines as worker nodes for clusters, in sizes you define centrally.desktop_windowsKubeVirt: VM managementvMetalEnterpriseExperimentalDeploy KubeVirt, create reusable VM templates, and manage virtual machines from the platform UI.computerKubeVirt: VMs as workloadsvClusterEnterpriseLet workloads in a cluster create KubeVirt virtual machines, reusing KubeVirt on the control plane cluster.terminalTerraform: nodes and networksvMetalFreeRun Terraform scripts with OpenTofu to create worker nodes and the infrastructure behind network environments.cloud_uploadS3-compatible OS image storevMetalEnterpriseStore uploaded OS images in AWS S3 or an S3-compatible store such as MinIO or SeaweedFS.auto_awesome_motionKarpentervCluster PlatformGive a cluster or a project its own Karpenter node pool on EKS, AKS, or GKE, and target node pools by cluster pod labels.
Networking
Connect servers and virtual machine nodes to tenant networks, automate the network for clusters, and add workloads to a service mesh.
AI and GPU workloads
Schedule GPU workloads, install tested AI platforms onto clusters as Certified Stacks, connect clusters that NVKE created, and serve and scale models inside a cluster.
layersNVIDIA DSX OSvCluster PlatformSee how NICo provisioning, GPU health tooling, and the bundled NVIDIA Certified Stacks fit together in the platform.inventory_2Certified StacksvCluster PlatformBrowse the bundled catalog, including Saturn Cloud Enterprise and NVIDIA stacks, and deploy tested multi-application stacks with dependency ordering and health gates.model_trainingNVIDIA Run:aivCluster PlatformDeploy NVIDIA Run:ai with a dedicated or central control plane for its tenants, from a Certified Stack.boltNVIDIA DynamovCluster PlatformInstall the NVIDIA Dynamo inference runtime into a cluster or control plane cluster, so it can serve models across GPU nodes.account_treeNVIDIA NVKE management planevCluster PlatformInstall the NVKE management plane into a cluster on private nodes from a signed release. NVKE is an upstream preview for evaluation.linkNVKE clustersvCluster PlatformConnect clusters that NVKE created to the platform, deploy applications into them, and configure metrics.scheduleNVIDIA KAI schedulervClusterSchedule GPU and CPU workloads in a cluster with queuing, fractional GPU allocation, and topology awareness.smart_toyInference serving stacksvClusterEnterpriseCompose a serving stack such as vLLM or KServe inside a cluster and expose an inference endpoint.trending_upGPU and inference autoscalingvClusterEnterpriseScale workloads with the Horizontal Pod Autoscaler (HPA) and KEDA, using NVIDIA DCGM GPU metrics and model-serving metrics.
GitOps, CI/CD, and automation
Create clusters and other platform resources from Git, a pipeline, infrastructure as code, or your own product's control plane.
Identity, secrets, and policy
Sign in through your identity provider, give workloads cloud identities, and bring in secrets, certificates, and policy engines.
loginSSO providersvCluster PlatformEnterpriseSign in with GitHub, GitLab, Google, Microsoft, Okta, SAML 2.0, LDAP, or any OIDC provider, and sync groups to teams.keyOIDC token exchangevCluster PlatformExchange short-lived OIDC tokens from GitHub Actions, Azure DevOps, and similar CI/CD runtimes for platform access keys.verified_userPlatform as an OIDC providervCluster PlatformFreeLet other services, such as a Harbor registry, sign users in through the platform.badgeEKS Pod IdentityvClusterFreeLet workloads access AWS services through an IAM role linked to their service account.badgeGKE Workload IdentityvClusterFreeLet workloads access Google Cloud services through a linked Google service account.badgeAKS Workload IdentityvClusterFreeLet workloads access Azure resources through a federated Microsoft Entra identity.lockHashiCorp VaultvCluster PlatformEnterpriseSync secrets from Vault into project secrets and global secrets, without copying them by hand.vpn_keyExternal Secrets OperatorvClusterEnterpriseSync secrets from an external store into a cluster, using the operator installed on the control plane cluster.verifiedcert-managervClusterEnterpriseIssue TLS certificates for cluster workloads from the cert-manager installation on the control plane cluster.policyKyverno storage policiesvCluster PlatformEnforce per-PVC size limits and storage class allowlists in every cluster created from a template.
Observability and fleet management
Collect metrics and logs across the fleet, watch GPU health, forward telemetry to your own backends, and manage clusters from the tools you use today.
Connectors and extensions
Store the connection to an external service once and grant tenants permission to use it, or extend vCluster where no integration exists.
Integrations by product
Each product's docs also has its own integrations overview. See vCluster integrations, which also shows the worker node models each integration supports, and vCluster Platform integrations.