Why Watchdock

I was tired of finding out last.

A letter from the backend engineer who built Watchdock, about eight years of being the on-call rotation.

For eight years I was the backend engineer, which on a small team means you are also the DevOps engineer, the on-call rotation and the person who explains the outage to everyone the next morning. Nobody hands you that second job. It arrives the first time a server falls over and you are the one who knows where it lives.

I have had the 3am page that turned out to be nothing, and I have had the worse one, where the incident was real and I heard about it from a customer an hour after it started, while every dashboard I owned was still green. The tools had all been watching. None of them had told me.

An alert that pages you for nothing is worse than no alert at all, because it teaches you to ignore the next one, and the next one is always the one that matters.

Five tabs and one incident

When something breaks, the answer is never in one place. The server's CPU is in one tool and the container that restarted is in another. The stack trace sits in a third, the access logs are in a terminal over SSH, and the uptime check that should have caught it lives in a dashboard you have not opened in a month. So during the worst hour of your week you become the integration, copying timestamps between tabs, lining graphs up by eye and pasting screenshots into Slack while customers wait.

Every one of those tools did its job. What nobody did was put the check that failed, the container behind it and the line of code that caused it on the same page, which is the only view that answers the question you actually have at 3am: what broke, and why.

Built for a team I did not have

Every platform I tried assumed a platform team behind it. Setup took weeks, each concern wanted its own agent, and the bill grew with every host I added, long before the product behind it made any money. I had a handful of servers, a Postgres database I cared about deeply, and no patience for a dashboard that needed a week of configuration before it told me anything useful.

So I built the one I wanted

Watchdock is one agent per server and one page per incident. Servers, containers, logs, uptime and errors arrive already joined together, so a failing check shows you the container crash-looping behind it and the error that started it. Alerts go to Slack, email or your phone and respect your channel instead of flooding it. When an error comes from your code, the AI reads the stack trace and the file, names the commit that introduced it and drafts a fix. You review it and approve it; nothing merges on its own.

Watchdock watches itself

I run Watchdock on Watchdock, and it has not always been graceful. Each of these happened in production, on the platform you would be trusting, and each one changed how the product works:

Database at 186% CPUA query I shipped, run on every requestBack to 2% the same day
Disk filling with logs76 GB of healthy-traffic logs nobody neededNow we keep errors, not noise
The platform went downThe monitor stopped and nothing said soIt now pages me when its own checks go quiet

I wrote up the database incident in full and published it, because if I am asking you to trust Watchdock with your infrastructure, you should be able to see how I handle mine.

Who it is for

Watchdock is for the engineer who is also the on-call rotation, the ops team and the person writing the incident update at 9am. If that is you, this is the tool I wish I had for the last eight years, and the one I now trust with my own sleep.

Welcome to Watchdock.

Nick LangatNick LangatFounder and engineer, Watchdock

Stop finding out last.

Free forever on Hobby. Upgrade when your infrastructure grows.

Start monitoring for free