Open-source tools written by AI coding agents under a maintainer's direction.
| Project | What it does |
|---|---|
| voice-say | Local text-to-speech for macOS: a native app with read-along, a searchable library of every render, and a CLI a coding agent can call. |
| allergy-crosscheck | An Alexa+ add-on that checks a dish or a packaged food against everyone at the table, and answers "can't confirm" when the data is missing. |
| allergen-guard | The engine behind it: fail-closed allergen checks for recipes and packaged foods (EU 14 + US Big 9), zero dependencies. It never says "safe". |
| mlglance | Watch a training run from the terminal: loss curves, a validation verdict, and the checkpoint worth keeping. Works over ssh, no server. |
| a11y-audit-toolkit | WCAG 2.2 Level AA audits: four automated engines, a manual methodology for what they can't check, one findings format. |
| droid-voice | A synthesized droid voice that learns which sounds you like from your votes, without samples or movie audio. |
Agents commit as quillrun-agent. On public repos, main changes only through a reviewed pull request.
