See the signal
Collect GPU and platform signals into one incident model with the evidence that triggered it.
Open source · Kubernetes native · Fail closed
GPU fleet reliability,
in your control.
KubeNeuron is an open-source control plane for finding GPU degradation, making remediation decisions visible, and building toward safe automation without surrendering your fleet to a black box.
Community Edition is available as source. Runtime automation remains gated by design.
fleet / status$ kubeneuronctl status
- installation
- gpu-platform
- mode
- DRY_RUN
- signal path
- OBSERVED
- operator
- READY
Guardrails first. Automation only when the fleet proves ready.
A reliability system, not a reflex
Make every GPU response
earned and explainable.
Review the path
Express a typed escalation path in versioned playbooks, approvals, maintenance windows, and audit events.
Earn automation
Start in DryRun, validate the environment, then decide when a workflow deserves a guarded action.
Community Edition
Start open.
Stay in control.
The core control plane is built in the open for platform teams who need inspectable policy, portable Kubernetes primitives, and a clear record of what the system did—or deliberately did not do.
Read the sourceOwn the policy
Typed playbooks live with your infrastructure.
Keep the evidence
Incidents, decisions, and transitions remain auditable.
Choose the pace
DryRun is the default; enablement is a safety milestone, not a toggle.
Built for the real operating model
The team that owns the cluster
owns the last word.
Policy, status, and ownership are first-class resources.
Work with CRDs and a reconciliation model that fit the way platform teams already run fleets.
No silent leap from a signal to a destructive action.
Decisions are explicit, reviewable, and recoverable.
What comes next
Enterprise,
when it earns its place.
We are designing a future Enterprise offering for teams that want a managed operating layer around the open-source control plane—not an opaque replacement for it.
Community$0
Open-source control plane
- Self-host in your Kubernetes environment
- Typed policy and playbook resources
- DryRun-first incident workflows
- Public source and roadmap
PLANNED
EnterpriseLater
A managed operating layer, shaped with design partners.
- Fleet-wide policy governance
- Operational evidence and reporting
- Guided rollout and support
- Potential hosted control-plane options
Clear by design
The important answers.
Can we use KubeNeuron today?+
You can evaluate and contribute to the Community Edition today. The current runtime is deliberately DryRun-first while hardware-gated remediation safety is completed.
Is the Enterprise offering available?+
Not yet. It is a direction, not a product promise or a published price. The open-source control plane remains the foundation.
Will a hosted SaaS replace self-hosting?+
No. Any future hosted option should reduce operational overhead while preserving portable policy, evidence, and operator control.
Build the safe path first