I open-sourced Assay today, the desktop tool my own consulting practice runs on.
It audits a local business's AI-readiness by reading their raw page source, six deterministic checks, and drafts the outreach packet: scorecard PDF, schema starter kit, delivery drafts. Then it stops, and nothing auto-sends. No sender is implemented in the codebase; each artifact needs individual approval, and even then the sending is mine to do by hand.
My whole pitch to a prospect is that every claim in the audit can be reproduced in their own browser with view-source. The constraint lives in the types and the tests.
The calibration set is two client scans, and reconciliation agrees on severity band rather than exact score. MIT licensed, source only for now. If you work in outreach tooling or structured data, I would value your read on the approval gate.
It all started a few months ago with my mom showing me her website for her local shop which also has a Shopify storefront. She'd been upset because she felt like she was losing out to a local competitor.
Me, being obsessed with data wanted to prove that wrong. The site was in terrible shape, I knew nothing about pagesource, cats.txt (lol) or how AI has been beginning to rule how search works.
This tool is the culmination of that work, and I'm happy to share it all to you.
Check it out before launch happens tomorrow at https://www.ryanlenk.com/pages/a...