I built it to do something new.
That is the honest version.
When we decided to add an AI layer to Quave ONE, I told the team: "let's build this, but it's not going to be helpful."
I was the loudest skeptic in the room, which is a funny position when you own the company hehe.
And yes, maybe there was some hype in the decision too.
One day into using it, with the MCP still running only on my machine in development mode, I caught myself not opening the logs page anymore. It was easier to ask from Cursor and get logs, metrics, deploys, and the likely problem in one answer.
One of our customers runs 192 environments on us.
You don't click through that every morning.
Nobody does.
Then I got greedy and created a watchdog for our own production. I wrote one prompt explaining what I wanted watched, the AI used our MCP tools to build the automation, and now it runs every day and pings Slack when something is off.
It already caught our jobs memory jumping from 48% to 73%.
I did not write a threshold.
I did not write an alert rule.
One conversation with AI, and now anomaly detection runs all the time.
The part that convinced the skeptic in me was not the "AI magic". It was the boring safety part, which is the important part: same permissions as the dashboard, same APIs, scopes enforced on the server, not by the agent, scoped keys, and every action audited.
The AI is an impersonation of you. It can never do something your own permissions do not allow.
And yes, our platform runs on our platform. The watchdog from the video is running on our real production account right now.
Today we are back on Product Hunt as Quave ONE, after 13 years as a team, 2,000+ apps in production, profitable and bootstrapped:
https://www.producthunt.com/products/zcloud-2?utmsource=quaveone&utmmedium=social
Quave ONE is the cloud platform you run by talking to it.
Come poke at it and tell me where it falls short. I prefer that over polite clapping.
See you there ;)