In the dark, catching bugs before they bite.
Every policy blocking your Kubernetes workloads — discovered, correlated, and explained to the developers who need it.
The Problem
NetworkPolicies, Gatekeeper constraints, Kyverno policies, Istio rules, resource quotas, webhooks. When something blocks a deploy, developers get cryptic errors and no answers.
Policies live in Cilium, Gatekeeper, Kyverno, Istio, native K8s, and custom CRDs. No single tool shows them all.
Developers see connection timed out or webhook denied the request with zero context
about which policy caused it.
Platform teams write policies. App teams hit them. The debugging loop burns hours and goodwill on both sides.
The Solution
Potoo watches every policy-related resource in your cluster. When something breaks, it tells developers exactly what happened and how to fix it.
Features
Potoo doesn't enforce anything. It discovers what's already there and makes it understandable.
Finds constraints from NetworkPolicy, Cilium, Gatekeeper, Kyverno, Istio, webhooks, and any custom CRD you register.
K8s Events on affected workloads, ConstraintReport CRDs, Slack integration, and workload annotations.
Multi-tenant safe. Developers see what they need to unblock themselves. Cross-namespace details stay hidden.
Built-in MCP server lets Claude, Copilot, and other AI assistants query constraints and explain errors directly.
Validate manifests before deploying. Know which policies would block your workload before kubectl apply.
Machine-readable reports per namespace. Query with kubectl, pipe to dashboards, or feed to automation.
How It Works
Potoo runs as a controller in your cluster. No sidecars, no agents on nodes, no policy changes.
Watches all policy CRDs in the cluster via dynamic informers. Adapters normalize each type into a common model.
Constraints are indexed by namespace, labels, and type. Lookups are instant regardless of cluster size.
When errors appear, Potoo matches Kubernetes events to the constraints that caused them.
Developers get privacy-scoped notifications with what went wrong, why, and how to fix it.
Get Started
Install via Helm, then use the CLI or let your AI assistant connect over MCP.