Taking requests

Bring the code.We'll host it.

Send us a repository, a paper's code, or a folder of scripts. If it's useful to others, we'll turn it into a service anyone can run, and you earn on every run.

01How it works

From a repo to a service anyone can run.

No Dockerfiles, API specs, or servers on your side.

  1. 01

    Send the source

    A link or the files: a repository, a paper's code, a notebook.

  2. 02

    We take a look

    Does it run, and would someone besides you call it? If not, we'll say why.

  3. 03

    We package it

    Inputs, outputs, the hardware it needs, and a price you agree to.

  4. 04

    It goes live

    Under your name, runnable from a browser, Python, or an agent.

Today a person does steps two and three. Soon an agent will, and you'll be able to watch it work.

02Send it over

Point us at it, or upload it.

We read every request ourselves. Not everything is a fit, and we'll tell you either way.

Good fits

  • Models people already email you to run
  • Code from a paper that's hard to install
  • Anything that takes a file or a few numbers and returns a result

A repository, a paper with code, or a project page.

If the trained model lives somewhere else, like Hugging Face.

Drop a zip, scripts, or a notebook here, up to 200 MB each.

I wrote itMy group didSomeone else did

03The terms

Your code, your name on it.

Hosting it doesn't make it ours.

Yours to keep
Your service page, your license. Take it down whenever you want.
Priced your way
Free, or a price per second of compute. You earn on every paid run.
Credit travels
Someone else's model goes up under their names, and stays free unless they agree.

Rather build it yourself?

A coding agent connected to Ouro can turn a repository into a service in an afternoon.