Show HN: Kubently – Debug Kubernetes Clusters Agentically (kubently.io)

Kubently is a new AI-driven tool for conversational debugging of Kubernetes clusters that combines real-time streaming, multi-LLM intelligence, and a lightweight per-cluster executor model. It exposes a horizontally scalable FastAPI-based API that coordinates cluster executors via Redis pub/sub and Server-Sent Events (SSE) for ~50ms command delivery and near-instant execution. Executors are read-only by default, honor configurable RBAC, and the system includes command validation, OAuth/OIDC and TLS support (cert-manager) for enterprise readiness. Integration points include a REST API, Node.js CLI, and an automated test framework. For the AI/ML and SRE communities the significance is twofold: it operationalizes agentic, multi-LLM troubleshooting across live environments, and it does so with a production-aware architecture (A2A agent-to-agent protocol, LLMFactory for multiple providers, tool-call interception and streaming responses). That enables systematic diagnostics, todo tracking, and multi-agent collaboration while supporting unlimited API pods via Redis-backed horizontal scaling. Practical implications include faster root-cause discovery and tighter CI/ops automation, though teams should still manage LLM risks (hallucinations, permissions) through read-only defaults, RBAC, validation, and human oversight. The project is open-source with docs, community support, and a quick-start path for deployment anywhere Kubernetes runs.