I totally get the concern giving an LLM 'write' access to a production cluster is definitely a scary thought.
To clarify: Kore is designed to be read-only by default. It’s an exploration and investigation tool, not a deployment or management engine. It uses the AI to help correlate data, explain complex status fields, and navigate the relationship between resources (like tracing why a Service isn't hitting its Endpoints).
Since the tool does not perform any apply, patch, or delete operations, there is no risk of the AI 'hallucinating' a destructive command that would actually impact the cluster infrastructure.