For API users, the promise is simple.
The first words show up quickly, the answer keeps streaming, and a traffic spike from someone else does not turn your app into a waiting room.
Shared API1101msPerchy191ms
AI APIs without traffic panic
FindYourPerch
Your AI feature should feel the same at 2 PM as it does at 2 AM. Perchy gives each active app a clear temporary lane, then releases it when the app goes quiet.
Play with traffic
Slide traffic up. Watch a shared API make everyone wait.
Perchy keeps your app in a clear lane on machines that are not filled past the point where people start feeling the slowdown.
Shared APIPerchy lane
Shared API1101ms69/sec stream
With Perchy191ms90/sec stream
Feels faster910mssaved before the first words
Your app21/32comfortable positions
App requests
Spare compute
The product in one picture
Every active app gets one clear position.
A normal shared API sells the same capacity again and again, so every user can feel slower when traffic rises. Perchy sells positions only up to the calm operating range of each group of machines.
When your app sends requests, the position is yours. Stop sending for the timeout you choose, and the position returns to the market.
Clear laneMeter pausesHosts earn
Pay for being present, not for a vague token pool.
You pay while your app occupies a position. The minimum is one second. A longer idle hold costs more, but keeps your lane ready between bursts.
Current burst$0.30
For GPU owners, it should feel like listing spare capacity.
Connect from home, the studio, or a small lab. No public address or network setup is required; your machine can earn while it is available.
Projected hour$3,570.74
Still just code
The interface can stay friendly because the product does the hard part.
A clear lane in code
const answer = await perchy.chat({
model: "fast",
lane: "keep-clear"
});Apps ask
Perchy matches
Clear positions
Spare compute earns