Codeball AI
AI code review that scores pull requests and auto-approves safe ones.
About
Codeball, from Sturdy (Y Combinator W21), is a code review AI trained on more than one million pull requests. It scores each PR from 0 (needs careful review) to 1 (merge it), labeling risky changes for human attention while auto-approving safe ones, with the goal of fast-tracking easy reviews so engineers focus on the tricky ones.
It installs as a GitHub Action in about two minutes with sane defaults and is fully customizable through the action configuration: approve pull requests, label on approval, or flag caution as needed. Codeball is used by over 100 open-source repositories and offers an online demo where users can run historical tests on GitHub repos.
What it can do
- 0-to-1 PR risk scoring
- Auto-approval of safe pull requests
- Caution labels for risky changes
- Online demo with historical repo tests
Setup
Add sturdy-dev/codeball-action@v2 to a GitHub Actions workflow on pull_request events.
Prerequisites
A GitHub repository with Actions enabled.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 104; Trending score 72/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.