io.github.inhuman/mcp-k8s-ephemeral-jobcloud

Kubernetes Ephemeral Job

Verified · today

Runs a command in a throwaway Kubernetes pod and returns exit code, output and artifacts.

Install

claude mcp add mcp-k8s-ephemeral-job -- docker run -i --rm docker.io/idconstruct/mcp-k8s-ephemeral-job:v0.7.1

Requires: MCP_K8S_ALLOWED_IMAGES, MCP_K8S_KUBECONFIG, MCP_K8S_NAMESPACE, MCP_K8S_TRANSPORT, MCP_K8S_DEFAULT_TIMEOUT_S, MCP_K8S_MAX_TIMEOUT_S, MCP_K8S_MAX_OUTPUT_BYTES, MCP_K8S_MAX_ARTIFACT_BYTES, MCP_K8S_MAX_CONCURRENT, MCP_K8S_DEFAULT_CPU, MCP_K8S_DEFAULT_MEMORY, MCP_K8S_SIDECAR_IMAGE, MCP_K8S_CLONE_IMAGE, MCP_K8S_CLONE_SECRET, MCP_K8S_CACHE_PVC, MCP_K8S_CACHE_MOUNT_PATH, MCP_K8S_JOB_EXTRA_ENV, MCP_K8S_AUTH_TOKEN

Our take

This server runs one-off commands inside temporary Kubernetes pods and returns output, exit status, and artifacts. It is for platform engineers, CI-style automation, or agent workflows that need isolated execution inside a cluster without maintaining long-lived workloads. The design is powerful and operationally relevant, but it carries obvious blast-radius risk if permissions are loose, so maturity should be judged largely by its RBAC and artifact-handling discipline.

reviewed by hand · 2026-07-25

Something wrong or dead here? Report it