Pi's tagline is "there are many coding agents, but this one is yours." That is the difference that matters when you run a Shopify catalog. You do not bend pi to fit a generic workflow. You configure it to know your voice, your field rules, your length limits, and your taxonomy, so every pass it makes on your products reflects decisions you made in advance, not defaults someone else shipped.
What stops most merchants from letting an agent loose on a live catalog is the storefront itself. It converts, and most ways of wiring AI to Shopify write straight to it. Scratch brings your catalog down as a folder of product files. Pi rewrites them on your machine, with the validators and voice guides you have set up. You read the result as a diff and approve what ships. Prices, variants, and inventory stay locked throughout, so the pass you cannot afford to get wrong simply cannot happen.
How it works
- Scratch pulls your catalog into files. Every product comes down as one JSON file in Shopify's GraphQL Admin shape, in a folder on your machine. Pi can read the whole folder at once.
- Pi edits the products. Point pi at the Scratch folder and describe the rewrite. Because pi is yours, it already knows your brand voice, your field length limits, and any validators you have wired in. It works the catalog file by file and reports the changes for review. It edits the files, never the live store.
- You review every diff and publish. Scratch puts each changed field next to the original, word by word. Approve what ships, and Scratch writes only those products back through the Admin API, one at a time.
What people use it for
The catalog work that stalls because the Shopify admin is one product per screen, and because generic AI tools do not know your rules:
- Rewrite every description to a consistent length and voice, with pi already knowing what that voice sounds like.
- Backfill SEO titles and meta descriptions across the catalog in a single pass.
- Open each description with the material, fit, or benefit instead of boilerplate, enforced by a validator you wrote once.
- Tidy vendor, type, and tags so collections and filters behave.
- Refresh product copy after a rebrand, with pi configured to the new guidelines from day one.
Why not an MCP server?
A Shopify MCP server or app wires pi straight to your live catalog. One wide pass ships to every product at once, and you spend the afternoon undoing it by hand. The configuration that makes pi yours, your validators, your prompts, your voice rules, still runs, but it runs against a live storefront with no gate between the rewrite and the customer.
Scratch lifts the publish step out and hands it to you. Pi can change anything in the local folder; only you can ship it. On a storefront that is taking orders, that gap is the whole point. And because pi is configurable at every layer, you can wire Scratch validators directly into its workflow so problems are caught before a diff even reaches you.
What pi edits in Shopify
- Product titles, descriptions (HTML), and handles
- Vendor, type, and tags
- Product SEO title and meta description
- Articles, blogs, and pages
Prices, variants, inventory, and metafields are locked at the connector. Pi cannot write them back even if a prompt asks it to, and length and taxonomy validators catch problems before a diff reaches you. The full field list lives on Scratch for Shopify.
Questions people ask
Is this an MCP server or a Shopify app?
Neither. An MCP or an app hands pi the publish button. Scratch keeps it. Pi gets the same full read and write access against a local copy, and publishing is a separate step you approve, one product at a time.
Can pi touch my prices or inventory?
No. Prices, variants, inventory, and metafields are locked at the connector level. They cannot be written back, even if a prompt asks for it.
How do I configure pi for my store's voice and rules?
Pi is designed to be configured. You write a prompt file or validator that describes your brand voice, field length rules, and taxonomy, then point pi at the Scratch folder with those configs loaded. Every pass it makes from that point on follows the rules you set.
Can it rewrite thousands of products from one instruction?
Yes, that is the use case. Run it on 50 products to feel the loop, then point it at the full catalog. Pi is built for large file-set jobs.
Can I roll a change back after it ships?
Yes. Scratch keeps the original beside the rewrite for every published product. You revert per row, and you decide which version stays.
See it on your own catalog
The fastest way to trust it is to watch it run on your products. See it run on your Shopify catalog, or download Scratch free and run the first pass yourself.