The story that sells itself
Every paying customer went through some version of this. Maybe you have too.
PagerDuty fires. High Latency — Payment Gateway. You roll out of bed, grab your laptop, and start the ritual you've done a hundred times before.
In the incident review, a colleague mentions they've been using an open-source MCP server that gives AI assistants like Claude or Amazon Q direct read-only access to the cluster — logs, metrics, K8s events, deployments, CloudTrail, all of it.
No dashboards. No tab-switching. You just ask a question in plain English and the AI queries everything at once.
The checkout pod keeps restarting in staging. You open Claude Desktop and type one question.
get_k8s_events call shows this has happened 3 times in the last hour, all triggered by the same request pattern. Recommended fix: increase the memory limit to 768Mi and add a payload size guard on the /checkout/process endpoint.The free tool is great — when you're awake and at your desk. But the worst incidents don't happen at 2 PM on a Tuesday. They happen at 3 AM on a Saturday, when nobody is sitting in front of Claude Desktop.
KubeOpsAI Agent closes that gap. It hooks directly into Alertmanager. The moment an alert fires, it runs the same investigation automatically — no human prompt needed — and posts an evidence-cited root-cause analysis to your Slack incident channel before PagerDuty even wakes anyone up.
