Skip to content
Technology

Stay up when others are down

Run your product, your AI inference and your internal platform active-active in a Global Virtual Cloud across AWS, GCP, Azure, neoclouds and your own servers. When a zone, a region or a whole provider fails, users are routed to the locations still serving. Workloads with at least two replicas in at least two locations are backed by a 99.999% SLA.
Uptime SLA

99.999%

reduction in cloud spend

30–50%

Use any service from AWS, GCP, or Azure

600+

Software, AI and infrastructure companies on Control Plane

Back the SLA in your enterprise contracts

Enterprise customers ask for an uptime number in the contract. On one provider, the region-level SLA you can lean on is 99.99%. On Control Plane your product serves from more than one provider at once, and workloads with at least two replicas in at least two locations are backed by a 99.999% SLA.

Multi-tenant products keep their routing. Customer domains, path rules and the services behind them deploy to every location from one spec.

Any infrastructureOne virtual cloud
KubernetesCloud accountsData centers

VMs · Bare Metal · Edge — Networking · Identity · TLS · Observability · Failover

Inference on whichever GPUs are up

Model serving can draw on hyperscaler GPUs, neoclouds and your own servers in the same Global Virtual Cloud. When one source fails or runs out of GPUs, requests go to a location that still has capacity.

The model APIs and data services you already use in AWS, GCP and Azure are reached with short-lived credentials through Universal Cloud Identity, from any location.

Deploy onceRun everywhere you choose
Your appGPU pool on LambdaEKS clusterAzure region

Runs in your own accounts · Credits and commitments still count

One platform for the clouds you acquired

Acquisitions arrive with their own AWS accounts, GCP projects and Azure subscriptions, each with its own consoles and runbooks. Existing clusters join Control Plane as locations through Bring Your Own Kubernetes, and every team gets the same network, identity model, policies and audit trail.

Moving a service from one provider to another is a change to a location list, so consolidating onto the providers you want doesn't wait on a migration project.

GVC · ProductionAI-native
Human operatorAI agentGPUsDataAI workloads
Identity · Policy · AuditSame guardrails for every operator

Sandboxes for agents, with real access and no keys

An agent needs to clone a repo, run the app and reach the services it depends on. Control Plane sandboxes are isolated environments inside your virtual cloud that are ready in 60 to 90 seconds and reach private services through Universal Cloud Identity and Cloud Wormhole, so the agent never sees a key.

Agents work through the MCP server with the same identity, policies and audit trail as an engineer.

Zero-trustEvery substrate
VMsBare metalKubernetes

Least-privilege access · Micro-segmentation · Encryption · Immutable audit trail

SOC 2 Type IIPCI DSS Level 1HIPAAGDPR

How a technology company stays up on Control Plane

Example: a SaaS company on AWS adds GCP and a GPU cloud

A software company serves its API from EKS in AWS us-east-1 and runs model inference on one provider's GPUs. Its enterprise customers want a higher uptime SLA in their contracts.

  1. EKS

    Bring the existing cluster in

    The EKS cluster joins Control Plane as a location through Bring Your Own Kubernetes. Nothing is rebuilt.
  2. EKSus-east4

    Add a second provider

    A Control Plane location on GCP us-east4 joins the same Global Virtual Cloud. The API deploys to both from one spec, and both serve.
  3. EKSus-east4GPU cloud

    Add a second GPU source

    A GPU location on a neocloud joins as well. The inference service runs there and in AWS, so losing one GPU source leaves the other serving.
  4. EKSus-east4GPU cloud

    Drill it

    Take AWS out of routing during a quiet hour. The API keeps answering from GCP, inference keeps running on the GPU cloud, and the audit trail records every step.
product-gvc.yaml

cpln apply --file product-gvc.yaml

kind: gvc
name: product
spec:
  staticPlacement:
    locationLinks:
      - //location/saas-eks-us-east-1  # existing EKS cluster
      - //location/gcp-us-east4
      - //location/gpu-neocloud  # GPU servers
    locationOptions:
      - locationLink: //location/saas-eks-us-east-1
        routingTier: 0
      - locationLink: //location/gcp-us-east4
        routingTier: 0  # both serve

Answers for your customers' security reviews

Your enterprise customers will send a questionnaire. These are the platform's answers. Your application's own controls stay yours.

  1. SOC 2 Type II

    Control Plane completed the AICPA SOC 2 Type II audit, performed by Prescient Assurance.
  2. PCI DSS Level 1

    Control Plane is a Level 1 PCI DSS compliant platform, assessed on site every year. A responsibility matrix shows which controls are ours and which are yours.
  3. Immutable audit trail

    Every action by people, service accounts and agents is recorded and attributable, across every provider in the Global Virtual Cloud.
  4. No stored keys

    Universal Cloud Identity issues short-lived credentials, so no workload, sandbox or agent holds a key that can be copied and reused.

Technology teams on Control Plane

“Control Plane enables us to deliver a SaaS product to enterprise teams across markets around the world, while at the same time maintaining our development momentum to roll out new features.”

See your product on two providers

Sign up free, or book a demo and we'll fail a location over on your stack.

Run Unbreakable
Apps and Agents

Sign up free. No cloud accounts required.