Core concepts

Catalyte models commerce with four primitives: workspaces, products, channels, and feeds. Once you have the picture, every feature slots in.

Workspace

The top-level container. Everything is scoped to a workspace: products, members, API keys, OAuth grants, channels, logs. Row-Level Security in the database enforces strict isolation between workspaces.

Product

A single sellable item, identified by a sku that is unique within your workspace. The shape is intentionally flat:

{
  "sku": "TSHIRT-BLACK-M",
  "title": "Heavyweight tee",
  "description": "12oz cotton, boxy fit.",
  "brand": "Atelier",
  "price_amount": 48.00,
  "price_currency": "USD",
  "inventory_quantity": 120,
  "image_url": "https://cdn.example.com/tee-black.jpg",
  "url": "https://shop.example.com/tee-black-m",
  "tags": ["staple", "unisex"]
}

Prices are decimal amounts in the listed currency (e.g. 48.00 USD = $48.00), stored at 2-decimal precision and serialized to each channel's expected format on the way out.

Channel

A destination for your products. Channels come in three shapes:

• Feed channels — Google, Meta, TikTok, Pinterest. They pull a URL.
• SFTP channels — ChatGPT (ACP), Amazon, eBay, custom. We push on a schedule.
• AI channels — the MCP server and the public directory. Agents read live.

Feed

The serialized view of your catalog for one channel. A Google feed is XML; a Meta feed is CSV; an ACP feed is JSON. The shape is fixed by the channel; the contents are your products, transformed to fit.

How data flows

1. You write to products via the UI, CSV import, REST, or MCP.
2. Catalyte stores them in Postgres with full row-level isolation.
3. Each enabled channel has a feed URL and/or an SFTP schedule.
4. External platforms pull, or Catalyte pushes, on their own clock.
5. Every event lands in Logs.