anyone actually using AI SREs / agents to debug k8s incidents in prod?
retained text
curious what people are running when stuff breaks. is it 1. an AI SRE that picks up the alert (holmesgpt, aws devops agent etc) 2. claude code / codex with plain kubectl 3. same thing but hooked up to an mcp server that gives it cluster context (k8sgpt's mcp, radar, ...) full disclosure I work on radar (it's open source). we tested this a bit with injected faults but that's not the same as a real outage what's worked for you and what hasn't? trying to figure out what we should improve on our end
document fetched 2026-10-09 00:40 · canonical https://reddit.com/r/kubernetes/comments/1x0wbf3/anyone_actually_using_ai_sres_agents_to_debug_k8s