Hey Product Hunt! Today I'm launching KubeClaw, an open-source Helm chart that deploys OpenClaw to Kubernetes with production-grade guardrails. OpenClaw is an incredible AI assistant, but the common skepticisms are real: How do I know what it's doing? What is it calling? Can I control what leaves my network? How do I make upgrades predictable instead of scary? KubeClaw addresses all of that by making every action observable and every boundary enforceable using the tools cloud-native engineers already trust: Kubernetes, Helm, and OpenTelemetry. It's OpenClaw for skeptics. OpenClaw for the cloud-native software engineer. Run it on a single-node cluster on your laptop, a spare computer under your desk, or on any cloud provider. One helm install gives you: - Durable storage so your data survives pod restarts - Egress DNS filtering with default-deny outbound, explicit allow/deny lists, and full query logging - Unified observability using Wide Events. - LLM proxy with per-agent API keys, budget caps, and model fallback routing - Gateway API routing for clean, single-hostname access to all services - Tailscale integration so you can expose it on your tailnet or SSH in without public ingress - Chromium sidecar for browser automation - Digest-pinned images so nothing drifts silently between upgrades - GitOps-friendly config that plays nice with your existing workflows It's Apache 2.0, fully open source, and ships as a single OCI chart. What ran, what changed, what it called, and what it emitted are all auditable. Would love your feedback. Happy to answer any questions here! - Michael

KubeClaw

Production-grade OpenClaw on Kubernetes

Production-grade OpenClaw on Kubernetes