Skip to main content
AI labs train and test coding agents on environments: a real repository at a real commit, a task someone actually asked for, and a way to check the result. The best source of those is work that already happened. payback records your coding-agent sessions on open-source repositories, turns the good ones into environments, licenses them to labs, and pays you for each one that sells. You keep working the way you do now: there are no tasks to pick up and nothing to annotate.

​
Install

macOS or Linux, with git. The script downloads one self-contained binary, links payback into ~/.local/bin, and runs setup. You can read the script before running it. Install has the details.

​
What happens next

1

Sign in with GitHub

Setup opens your browser once. payback asks GitHub for your profile and email only; it never gets access to your repositories.
2

Your past sessions are offered for review

Setup finds sessions your coding agents already saved, on repositories that are public on GitHub, and offers to upload them. Nothing is uploaded until you say yes.
3

Passive income is turned on

A small hook goes into each coding agent on your machine. From then on, each turn in a public GitHub repository is recorded and uploaded in the background. Which repositories are recorded explains the rules, and they are yours to change.
4

Sessions are reviewed, built, listed and licensed

Every uploaded session is reviewed. The ones that become environments are offered to labs, and each one that is licensed pays you. Where a session goes and Getting paid cover the rest.
Supported agents: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot CLI, Cline, Pi and OpenCode, with aider for past sessions. See Agents.

​
Where to go from here

Which repositories are recorded

The policy, the allow list, exclusions, and the one command that changes them.

What is never recorded

Secrets, private repositories, anything outside a repository.

CLI reference

Every command, with its exact usage.

Your data

What lives on your machine, what we keep, and how to delete it.