The problem
Kubernetes access is either too open or too slow.
How it works
Read freely. Write with guardrails. No agents on the cluster.
Hoop sits at the network layer. No sidecars, no admission controllers, no cluster agents. The engineer or agent connects through the gateway using their OIDC identity. Every session is recorded. Every command is logged with identity, timestamp, and outcome.
get, describe, logs, topRead operations pass through without friction. Secrets and sensitive configmap values are masked in responses before they reach the engineer or agent.
apply, patch, scaleWrite operations are evaluated against guardrail rules. Low-risk changes pass through. Production deployments and configuration changes route for human approval.
delete namespace, drain node, delete pvcDestructive operations are blocked outright or require explicit, named approval from a designated approver. Not just any approval, but the right approver for the right resource.
Works with any Kubernetes distribution. No cluster-side installation required. Authentication via OIDC with short-lived tokens.
NATIVE TOOL SUPPORT
Lens, kubectl, helm. Your tools, invisible controls.
Your team keeps using Lens for pod management and kubectl for everything else. Hoop masks sensitive data in log output, blocks destructive commands at the protocol layer, and records every session. No cluster agents. No sidecars.
ORGANIZATIONAL IMPACT
From cluster governance to compliance evidence.
Every kubectl command, every pod operation flows through the gateway. The result: continuous compliance evidence and organizational risk metrics your security team can report on.
What happened in your clusters last month?
Most teams can’t answer that question. Hoop gives you a full session log from day one, without touching a single cluster.