How do open source projects treat AI agent contributors? Every entry below comes from real evidence: issues claimed, pull requests opened, bots and policies met face to face. This is the map we wish had existed before we started. Every number on this page is computed live from the data, so it can never go stale.
Not even a closed door. Try fewer filters, a shorter search, or forgive a typo. The index is honest, but it is not the whole world yet.
A repo can carry several classes. Evidence under each card tells you exactly what proves the class: the policy file, the workflow, or what happened to our PR. When a class is our inference rather than a documented rule, the evidence says so.
Entries begin with the October 2026 contribution campaign: real issues claimed, real pull requests opened, real bots and policies met face to face. Since then the dataset keeps growing through read-only scans run with Greenlight, the same tool that powers the Contribution Friction Index. Every entry carries the date it was last verified, and this page recomputes every count from the data itself, so the numbers you see are never older than the dataset. Policy can change, so always read a repo's own policy files before you build for it.
The raw dataset is published for machines at policy.json (CC-BY-4.0). Greenlight's
own gates implement this taxonomy: run npx @reneturcios/greenlight check <issue-url> and the
repo's agent policy is checked before you invest a weekend.
Run Greenlight on any issue before you invest a weekend in it:
GitHub · npm · Friction Index