2026 – present
Follow Apparel catalog sync
Agents sync Printful products and catalog metadata into Shopify for Follow Apparel, using supplier specs only, and a person approves anything that publishes.
Private repo, walkthrough on request
AI details
What the system actually does
Models
- AI agents do the Printful-to-Shopify sync.
Techniques
- Printful-to-Shopify product sync, including catalog metadata.
- Size charts and specs are pulled from supplier data.
- Brand locks: no invented products, no invented measurements.
- A person approves publish, and a listing isn't done until it passes QA.
Data
- Printful and other supplier catalog data for FollowApparel.com.
- The same supplier-data rule applies to the Amazon sales channel.
Evaluation
- Drafts are reviewed before publish.
- Storefront QA checks the pages involved, with Playwright in the tool set for browser QA.
- Multi-agent QA sits in front of anything that ships.
Challenges
- Size charts and specs are easy to get wrong when they are copied by hand.
- An agent must not invent a product or a measurement to fill a gap.
- Publishing stays with a person even when the sync itself is automated.
Innovations
- Catalog facts move from the supplier source into Shopify instead of being retyped.
- The content lock is the same rule on Shopify and on the Amazon channel.
Stack and tags
- Shopify
- Printful API
- Amazon
- Playwright
- ecommerce
- approvals
- catalog
Problem
Follow Apparel is a faith-based premium apparel brand on Shopify, with Printful print-on-demand fulfillment and an Amazon sales channel. Catalog details like size charts and specs were easy to get wrong by hand.
Why I built it
Product and metadata updates needed a path from the supplier source into Shopify that doesn't hand-copy specs or invent products or measurements.
How it works
AI agents automate the Printful-to-Shopify product sync and catalog metadata, pulling size charts and specs from supplier data. Brand and content locks hold the agents to those rules: no invented products, no invented measurements, and specs only from the supplier's data. The same supplier-data rule applies to Amazon catalog facts.
Publishing is gated. A person approves anything that publishes, and drafts are reviewed first. A QA pass, including storefront QA, has to happen before a listing counts as done. Playwright is available for browser checks. This work sits inside the agent fleet, so send, publish, and spend still stop for a person.
Outcome
Catalog updates can move from Printful and the supplier data into Shopify, and nothing goes live until it's been reviewed and checked.
An exception desk that flags drift (missing variant, spec that doesn't match supplier data, media order, unpublished versus active) and writes a page-by-page Playwright report is on the November 11, 2026 project list. I haven't built it yet.