In beta, with limited placesOpen to everyone in Q1 2027

Help Log in

↑↓ to move · Enter to open · Esc to close

Features · Write

A catalog a machine can reason about.

A product page is written for someone who can see the photo. An assistant composing an answer can't. Enrichment drafts the part nobody writes down, for every product, category, collection and content page in the store.

Book a demo → See what it writes
A small amber glass bottle on the open, lined page of a notebook.
What comes back

An identity and a reasoning set, not a blurb.

One click for one entity, a batch, or the whole catalog. The same depth on all four entity types, on both platforms. The model that ran is named on the entity.

  • A core identity: what the product is, in the terms an assistant needs
  • Three reasoning vectors: who it is recommended for, how the decision is made, which objection to answer
  • Quality scores and price positioning, marked as inferred
  • Technical details, as their own field rather than a paragraph
  • A meta description, capped at 160 characters
  • 12 to 18 search keywords, in the language your catalog is written in
  • A vector, so the catalog can be searched by meaning
  • Sector-aware: photo and video accessories get load capacity, heights, plate standards and light output cited with their units
Three cards on a small round table: one white, one pale and one dark.
What you teach it

Thirteen questions the photo can't answer.

Who it is for, what it is used for, what it is made of. How it is positioned, when it sells, where the price sits. What you can prove: warranty, reviews, certifications. Short answers are enough, skipped ones are fine, and the answers travel with every run.

  • Question sets per entity type, so a category is asked category questions
  • A Claude skill drafts the thirteen from the product's own URL, for you to approve and paste
  • The attributes ChatGPT and Gemini quote for the product that beat yours, added to the next enrichment in one click, from AI visibility
An abstract field of violet bars, used where there is no photograph yet.
Whose voice

Your voice, your prompt, your model.

Four layers travel with every prompt, and all four are yours to edit.

  • A brand sheet of up to 10,000 characters, typed or generated from the documents you upload
  • The most relevant passages of those documents, attached per entity
  • Your defaults: brand name, market, the meta-title format, a keyword blacklist
  • The prompt itself, per entity type, with a temperature slider
  • A dry run against a real entity that persists nothing and costs one enrichment
  • Ollama, OpenAI, Anthropic or Gemini, with your own API key
A close view of a typewriter's key levers and type bars.
After the draft

Read it, push it, or roll it back.

01

Read the diff

Every draft sits beside what the store serves today. Refine one signal with an instruction in plain words, or edit it by hand.

02

Push, or don't

Nothing reaches the storefront until you press Push: per signal, per entity, per batch, or all at once.

Where it lands →
03

Roll back

The last three versions are kept per entity. Restore one and it becomes a new version, so the restore is reversible too.

Before you run it

What people ask before the first enrichment.

Will it rewrite my page titles?

Products get a meta title from a template you control: brand, then name, then SKU by default. The Publish panel shows the exact value a push writes, with the model's suggestion beside it. Categories, collections and content pages keep the titles you wrote unless you switch that on in the store settings, and it is off by default.

Does it translate?

No. Keywords and every other signal come back in the language your catalog is written in. One catalog, one language.

Can it run on its own overnight?

No, and nothing is pushed on its own either. Every enrichment starts from a click and every signal reaches the storefront when you press Push. The two things that do run on their own only read: a sync when the catalog is stale, and a re-check of what was already verified.

What if I don't like what it wrote?

Edit the signal, or refine it with one instruction and run it again. If you already pushed, restore an earlier version inside Clione and push that one next. The storefront changes then, not before.

What does it cost?

Nothing, during the beta. Clione is taking selected clients into this stage, in limited places, free of charge, and prices are published when the beta ends in Q1 2027. A free trial exists as well: it connects a store, syncs it and enriches within its limits, and pushes nothing to the storefront.

Keep reading

Where it goes from here.

Publish

Schema and JSON-LD publishing

One graph per product page, through the platform's own API. Where the draft lands.
Write

Question manager

FAQs generated, imported, deduplicated and pushed as FAQPage, across the whole store.
Check

Verify

The live page read back, signal by signal, and an alert when one goes missing.
Start

Bring one product. We'll draft it while you watch.

Twenty minutes on a call: your store, one entity, what it serves today beside what Clione would write.

Book a demo → Read the graph it publishes
Clione is in beta: selected clients, limited places, free until the beta ends in Q1 2027.