Opening the library
The best prompts in the world are sitting in Notion pages and Slack threads, invisible to everyone except the person who wrote them. That's a waste — so I built a place where proven prompts come with receipts attached, and the ones that actually ship get ranked.
Oort started from a frustration I kept hitting: every time I needed a good prompt — for a cold email, a bug triage, a tweet — I'd end up re-solving a problem someone had already solved. Usually myself, six months earlier, in a file I couldn't find.
The existing "prompt marketplaces" were full of untested submissions ranked by upvotes from people who had never actually run them. What I wanted was different: every prompt should come with the number of times real users ran it this week, who authored it, and — when we could get it — a case study of what happened when someone shipped it into their real workflow.
That's the product. Bring your own API keys. Run whatever you want. The prompts that rise to the top are the ones people actually keep coming back to. The library is free; subscriptions live on the Flows workspace. Oort takes no cut of your API spend and refuses to pool keys — that breaks every provider's terms.
At the far edge of the solar system there is a reservoir of a trillion quiet objects: the Oort cloud. No one has ever seen it directly — we only know it exists because comets keep arriving, and their trajectories all point back to the same place. A comet is cloud-matter set in motion; it grows a visible tail only once it moves.
That is the whole product in one picture. Oort is the reservoir: every prompt stored, attributed, honestly at zero until someone runs it. Flows is what sets them moving. And like the astronomers, we don't rank prompts by what they claim to be — only by the trails they leave.
We will never ship a Verify Gate the AI can pass for you. The complete button unlocks when you've checked the work — self-graded AI is how 'looks done but is fake' happens in the first place.
Rankings derive from unique-user run counts, weighted by quality ratings. One whale can't rig the leaderboard. Neither can a voting brigade.
A case study without numbers isn't a case study. Every outcome we feature ties to a specific prompt or workflow that can be reproduced by someone who reads the piece.
We don't pool keys, don't train on your prompts, don't log your outputs for BYOK traffic. Provider terms respected — every time.
The goal isn't to impress anyone with our UI. It's to get you from 'I have a task' to 'I have an output' in under 90 seconds. Everything else is noise.
Oort and Flows are built by one person (Qira LLC), plus the open-source community the library mirrors — every imported prompt carries its source and license. The stats you see are real, which early on means many of them are zero. The honesty rule applies to us first.
It's a one-person shop and not hiring, but I love hearing from people who want to argue about how the product should work.