← back
Lobster Trap: OpenClaw in Containers from Local to K8s and Back — Sally Ann O'Malley, Red Hat
Takeaway
Container packaging gives agent deployments familiar tools for isolation, portability, secret handling, and state recovery.
Summary
- O’Malley runs OpenClaw in containers for reproducible environments, portability, explicit host access, and easier cleanup.
- Podman secrets combined with OpenClaw secret references separate API keys from ordinary configuration; Kubernetes offers corresponding deployment primitives.
- Mounted directories provide tools, skills, and MCP servers, while persistent volumes support backups and recovery.
- The same containerized application can move from local development to OpenShift or Kubernetes for broader deployment.
openclawcontainerskubernetes
Original description
Sharing a good agent setup usually means handing someone a pile of markdown, config files, and YAML and hoping they reproduce what you have. The answer in this demo is a container image: spin up a sub agent in two seconds from a Podman command, flip a flag for Kubernetes, and your personal setup becomes the team baseline. The stack is Podman locally, Kubernetes for distribution, same container image throughout. Secrets get two layers: Podman secrets for API keys on the host, OpenClaw secret refs inside the container. Volumes handle backup and recovery. An Nvidia team runs the same pattern in production with ten engineers each running their own OpenClaw in Kubernetes for model evals, doing work that used to take six people. Speaker info: / sally-ann-omalley Timestamps: 0:00 Introduction and background on Sally Ann O'Malley 1:25 Discovering and experimenting with OpenClaw 2:35 Benefits of running AI agents in containers 3:05 Introducing Forever Claw and sub-agents 5:52 Using containers for agent configuration and tools 6:21 Managing secrets with Podman and Kubernetes 8:10 Scaling agent workloads with Kubernetes 9:15 Nvidia team case study: Model evaluations 11:09 Backup, recovery, and persistence with volumes 11:47 Vision for workplace agent standardization 14:14 Local demo: Running OpenClaw with Podman 16:45 Choosing providers and configuring settings 17:50 SSH sandbox features 18:22 Running the Podman command and checking agent status 20:52 Transitioning agent workloads to Kubernetes and OpenShift