A unified DevOps command center that wires Claude directly into your production infrastructure. Bundles six providers under one MCP server: query pods and deployments via Kubernetes client-go, inspect ArgoCD sync diffs and rollback history, run PromQL queries and check firing alerts, pull PagerDuty incidents and on-call schedules, search Loki logs with LogQL, and manage Helm releases. Ships read-only tools for safe exploration (pod logs, resource usage, event streams) plus mutation and destructive tiers with dry-run and confirmation gates. Configured entirely through environment variables, so you can enable only the providers you actually use. Install with npx, point it at your kubeconfig and API tokens, then ask Claude to debug CrashLoopBackOffs or correlate metrics across your entire stack without leaving the chat.
claude mcp add --transport stdio notharshhaa-devops-mcp uvx devops-mcp