Does it use live stock to recommend?

No. It uses the stock status your catalog had when you loaded it, and you update it from the dashboard. If a product is marked out of stock, it flags it on the card instead of hiding it.

Can it compare products or cross-sell?

It can propose two options and explain how they differ with what it knows about each, and suggest a set or a complement if it's in your catalog. There is no "people who bought this also bought" engine: it recommends from what the shopper says, not from purchase histories.

What if the shopper asks for something I don't sell?

It says so in the first sentence and, if there's a real alternative in your catalog, offers it without pushing. It doesn't search the web or other stores.

Can I tell it what to recommend and what not to?

Yes. Your instructions override its method: if you don't want it recommending sets, talking about certain products or touching certain topics, you say so in the dashboard.

Does it recommend in the shopper's language?

Yes, it replies in the language it's written to, in six languages. Your product names and descriptions show exactly as they are in your catalog.

How does it learn my catalog?

For the demo, by reading the public information at the URL you give us. After that, from the dashboard: URLs, sitemap, PDF, CSV, your FAQ, or Shopify's products.json. You decide what it knows and update it as it changes.

Can my team take over?

Yes. Every conversation stays in the team inbox. When the agent should stop, or whenever you want, a person continues from there with the full history in front of them. The shopper doesn't switch channels or repeat anything.

What happens to my customers' data?

Each store runs its agent in an isolated environment. Conversations aren't shared with other customers or used to train models. History is kept for 30 days by default, adjustable between 1 and 90, then deleted automatically.

AI product recommendations

Recommend what you sell, with a reason

Q: Can it make up a product or a price?

No. It can only pick from the products in your catalog, and the name, price, photo and link come from your store, not from the model. If someone asks for something you don't sell, it says so.

It listens to what the shopper is after, asks just enough and proposes one or two products from your catalog with the why. It cannot make up a product or a price: it only picks from yours, and the name, price and photo come from your store.

Try it with your store

Real catalogControlled routesHuman handoffFlat monthly pricing

A “what would you recommend?” with no concrete answer is a sale going cold

Whoever asks wants a name, a price and a reason, not the whole list or a generic reply. The agent proposes one or two products with the reason they fit, and asks the minimum it needs to get it right. Like this:

How it recommends when the shopper doesn't know where to start

A conversation in VERDA 21, our demo store, opening with a vague request: the agent doesn't dump the catalog, it orients and asks one thing.

Hi, I want to start taking care of my skin but I don't know where to begin.

The simplest way is a three-step routine: cleanse, protect in the morning and one treatment. Does your skin lean dry or oily?

Oily, with the odd breakout.

Then: MAREA 01 to cleanse without tightness, CALMA 02 in the morning to balance sebum, and DUNA 05 as sun protection with an invisible finish. If you'd rather have the set, TRÍADA 21 brings cleanser, antioxidant and sunscreen for 82 €.

MAREA 0122,00 €

CALMA 0234,00 €

DUNA 0532,00 €

And will this get rid of my breakouts?

I can't promise that: it helps balance sebum, but it doesn't treat acne. If it's persistent, a dermatologist is the better call. Shall I open CALMA 02 for you?

Open CALMA 02See TRÍADA 21Talk to a person

One question per message, and at most two before naming a product. It says what each product is for and also what it isn't, and never promises results that don't come from your catalog.

The method is its own; the rules are yours. If your store doesn't want it recommending sets, or talking about certain products, you say so in the dashboard and your instruction wins.

What changes

Turn a product question into a useful next step

Your catalog only

It picks from the products you loaded. Name, price, photo and link come from your store, not from the model, so it cannot make up a product or a price.

Asks just enough

One question per message and at most two before proposing anything. If the request is specific, it answers directly; if it's vague, it orients in three sentences and asks.

Explains the fit, and the misfit

It connects what it knows about each product with what the shopper wants, and says what it isn't for too. No invented urgency, no discounts on its own.

From recommendation to product page

Every card leads to the real product page, with its price. If a product is marked out of stock in your catalog, it says so instead of hiding it.

What it uses to recommend, and what it doesn't

Three sources, and none of them is the internet.

Your catalog as you load it: name, price, description, category, photo, link and whether it's in stock. The better the description, the better the recommendation.
What the shopper says in the conversation, from start to finish: skin type, budget, who it's for. They don't have to repeat themselves.
The page they're on. If they open a product page, the agent knows, greets them about that product and can show them others from your catalog.
No live stock, no orders. Stock status is whatever your catalog said when you loaded it, and you update it from the dashboard.
No web search, no other stores. If someone asks for something you don't sell, it says so in the first sentence and offers the real alternative if there is one.

No surprises

What happens when you request the demo

Your existing site works as is. We don't ask for access or require any install for the demo.
You paste your store's URL
Try it with your store
https://your-store.com
We read your store's public pages: products, categories, shipping, and FAQs.
We build the agent on your catalog
Catalog
MAREA 01 cleanser
NUBE 04 ceramides
As soon as it's ready, you get an email with the link. You go in, ask it what a customer would, and decide.
We email you
Flatzer
See the agent in action with your store
https://your-store.com

The shopping path

From an uncertain need to the right product page

Flatzer keeps one shopping context across questions, recommendations, website movement, and human help.

The shopper explains the need in their own words, even without knowing a product name.
Visitor
“dry, sensitive skin”
“something for winter”
The agent asks for the constraints that change the recommendation.
Agent
redness or tightness?
do you use SPF daily?
It selects specific catalog products and explains why they fit.
Catalog
MAREA 01 cleanser
NUBE 04 ceramides
It opens an approved product or collection route and can perform configured page actions.
Approved route
/product/nube-04
action: click · fill
When the agent should stop, a person continues with the conversation context intact.
Inbox
full history
your team continues

What it needs from your catalog

Descriptions that say who it's for. "For sensitive skin", "fragrance-free", "for daily use": that's what lets it explain the fit and the misfit.
A price and a photo on every product. Without a price, the card shows none; without a photo, a blank. It invents neither.
One address per product. That's where the card leads. Without it, the card shows but doesn't link.
Stock status, if you publish it. It reads it when the catalog loads and marks "Out of stock" where it applies.
A deliberate order. Cards come from the first 40 products in your configured catalog: put the ones you want recommended up front.

Versus a support chat or a loose ChatGPT tab

What changes isn't that it talks: it's what it knows, where it lives, and what it costs.

	Support chat	Loose ChatGPT	Flatzer
Where it answers from	Your FAQ and macros	Whatever it knows from the internet	Your catalog, exactly as you loaded it
Recommends products	No: hands off to a person	Yes, but from any store	Yes, only from your catalog
Takes them to the page	No	No: the shopper is outside your site	Yes, through routes you approve
Hands off to your team	Yes, that's its job	No	Yes, with full context in the inbox
Price	Per resolved conversation or per seat	Free, but outside your store	Flat monthly fee, no sales commission

Where it answers from

Support chatYour FAQ and macros

Loose ChatGPTWhatever it knows from the internet

FlatzerYour catalog, exactly as you loaded it

Recommends products

Support chatNo: hands off to a person

Loose ChatGPTYes, but from any store

FlatzerYes, only from your catalog

Takes them to the page

Support chatNo

Loose ChatGPTNo: the shopper is outside your site

FlatzerYes, through routes you approve

Hands off to your team

Support chatYes, that's its job

Loose ChatGPTNo

FlatzerYes, with full context in the inbox

Price

Support chatPer resolved conversation or per seat

Loose ChatGPTFree, but outside your store

FlatzerFlat monthly fee, no sales commission

A support chat still earns its keep on incidents, and ChatGPT on general questions. Flatzer covers the stretch between the question and the product page.

Detailed comparisons: Tidio · Intercom · Crisp · Chatbase · ChatGPT

Flat pricing

One flat monthly price, no sales commission

You pay a monthly fee and know the amount before you start. No commission on sales, no charge per conversation or seat, and no extra fee for the AI.

Starter

$52.95/mo

To start on one channel

1,000 messages a month
Web widget + 1 channel

Growth

$106.95/mo

For stores with steady traffic

5,000 messages a month
Web widget + 2 channels

Pro

$193.95/mo

For high volume and more channels

20,000 messages a month
Web widget + 4 channels

See every plan and what it includes

Product truth

What it does today, and what we don't promise

We separate what the widget does today from the integrations that aren't there yet, so you know what you're testing.

Available now

Answers and recommendations grounded in your catalog
Approved routes and configured page actions
History, inbox, and human handoff
Isolated environment per store; history kept for 1 to 90 days

Not promised

Completing checkout or adding to cart in any store
Checking stock, orders, shipping, or returns in your system
A guaranteed increase in sales or revenue attributed to the agent

Questions ecommerce teams ask

Try it with your website

See the agent in action with your store

Paste your site's URL and we'll prep a demo on your real catalog.