Catalog AI
Platform · Attribute Sets

You write the brief. The agent follows it.

An Attribute Set is the specification for a finished record: which fields it must carry, what shape they take, and the instruction behind each one. It is the difference between generic output and output that fits your catalog.

Start freeBook a demo

Attribute set · Industrial fasteners

6 attributes · applies to 2 categories

DescriptionHTMLBuyer-facing copy. No specs, no pricing, no claims.
SpecificationsHTMLFactual only. Metric first, imperial in parentheses.
Thread pitchTextIn mm. State coarse or fine. Leave blank if unthreaded.
Tensile classTextISO 898-1 property class. Cite the standard.
Meta titleText · 55–75 chars
Meta descText · 150–160 chars
Why this page exists

“Our category is too specialised for this.”

It is the most common objection, and it is usually right about the tools it has been aimed at. Generic content generation produces generic content, and a catalog of hydraulic fittings does not survive contact with it.

The fix is not a better model. It is a better brief. An agent told to state thread pitch in millimetres, cite ISO 898-1 for the property class, and leave the field blank rather than guess will behave completely differently from one told to "describe the product".

Attribute Sets are where you write that brief, once, for each part of the catalog that needs its own.

Anatomy

What an attribute is made of

Name
What the field is called in your catalog and in every export.
Type
HTML for long-form copy, text for short values, and the other shapes your template expects.
Prompt
The instruction the agent follows for this field, on this kind of product. This is the part that does the work.
Scope
Applied per category and per brand, so a fastener is briefed differently from a candle without maintaining two accounts.
An example

The same product, two briefs

Both prompts are legitimate. They produce very different catalogs.

Generic

A weak brief

“Write a product description.”Produces something readable, unspecific, and interchangeable with every competitor running the same instruction.

Yours

A real brief

“One intro paragraph, then labelled sections with bullets. Address who it is for, how it is used, compatibility, and safety. Never state pricing, shipping, warranty, or medical claims. No technical specifications — those belong in their own field.”Produces something that fits your template, answers the questions your support inbox actually receives, and stays out of the fields that own that information.
Attribute sets

Questions

Do we have to build these before we can start?
No. A working default set is created with your organization — description, specifications, meta title, meta description — so a first job can run in minutes. Editing comes when the defaults stop fitting.
How specific can a prompt be?
As specific as the category deserves. Naming the units, the standard to cite, the fields to leave out, and the things never to claim are all fair game, and all of them measurably improve the output.
Can different brands be treated differently?
Yes. Sets attach per brand as well as per category, which matters when one supplier publishes rigorous datasheets and another publishes a photo of a box.
Do brands and categories have their own sets?
They do. Brand attribute sets and category attribute sets work the same way, so a brand profile and a category page each have their own specification.
What happens when we change a prompt?
Nothing, until you re-run. Existing values stay as they are; the new brief applies to the next job. Product history keeps both, so you can see what a prompt change actually did.

Write one prompt. Run fifty products. Read them.

This is the fastest way to find out whether Catalog AI can handle your catalog, and it costs a fraction of the free credits.

Start freeBook a demo

5 credits on signup · no card required