Craig McLuckie and Joe Beda, two of Kubernetes' original creators, are now applying their infrastructure expertise to agentic AI through their startup Stacklok. Their core argument: enterprise AI adoption is blocked not by model quality but by accountability, governance, and observability gaps. Drawing a direct parallel to how Kubernetes tamed container chaos, they're building Kubernetes-native, self-hosted infrastructure for MCP-based AI agents — with policy, identity, isolation, and auditability built in. The thesis is that enterprises need agents to be governable, portable, and 'boring' enough to trust, and that a protocol like MCP alone isn't sufficient without a full control plane around it.
Sort: