Observatory · v0.4 · public preview
An observability tool for the engineers who already know what they're doing.
Observatory is open source. We built it because the dashboards we were paying for kept asking us to look at the wrong things. It is not for everyone. It might be for you.
What it is, in plain terms
Observatory is a single binary that ingests metrics, logs, and traces from your infrastructure, stores them in a format you can actually query, and gets out of your way. It runs on one machine until you outgrow it, then it runs on more. There is no agent fleet to manage, no proprietary query language to learn, no sales call to schedule before you can try it.
We wrote it because the three of us, coming from teams of very different sizes, kept hitting the same wall. The hosted vendors were generous with their dashboards and stingy with their cardinality. The self-hosted stacks asked us to learn a new language every six months. Observatory is what we wanted to use ourselves.
It is not a replacement for your incident response process. It is not going to make a junior engineer into a senior one. It will not write your runbooks for you, and it does not run on vibes or AI. We mention this only because some of our peers in the category do, and we want to be plain about what we are and aren't.
Who we built it for
You run production systems and you have opinions about them. You've been on call long enough to have a feeling about which of your dashboards lie and which ones tell the truth. You'd rather read a query than click a wizard. You can tell the difference between a tool that respects your time and one that performs for your boss's boss, and you have limited patience for the latter.
If that sounds like you, this is probably worth thirty minutes of your time. If it doesn't (if you want a dashboard your CFO can read at a glance and a vendor who will hold your hand through procurement), there are good products for that and we'll happily point you at them. Observatory is not it.
A note on how we work
Observatory is open source under Apache 2.0. We accept contributions, we read every issue, and we publish a roadmap we actually update. The three of us pay for development out of pocket and through a small Open Collective. There is no enterprise tier waiting in the wings, and there isn't going to be. If we ever need to charge for something, we'll tell you on this page first.
We are not interested in being the biggest observability company in the world. We are interested in building a tool we would still recommend to a friend ten years from now, and we think those two ambitions are quietly incompatible.