It answers your visitors. Then it tells you why it couldn’t.
Every plan, Starter included, runs on a server set up for your store alone — its own database, its own disk, its own login. Your shoppers’ conversations are stored there, never in a database shared with other stores. That is what the price buys. On top of it, every question your catalog couldn’t answer lands in your console — and on Pro, which product records are too thin to sell, what’s missing, and what it’s costing you. From $99 a month. Monthly, no annual contract, no per-seat fees, no setup fee.
Starter
For stores putting a grounded assistant in front of shoppers for the first time.Runs onYour own server, database and login — shared with no other store.
- Sized for around 1,000 conversations a month · up to about 500 products · 100 pages of your site indexed
- The buyer agent on your site, answering from your catalog with our fast model
- Lead capture — every contact in your console
- Hand-off when it can’t answer — your contact page or a lead form, straight into your console
- Funnel analytics: sessions, engaged, cart, checkout
- Knowledge gaps — the questions your catalog couldn’t answer, last 7 days
- Email support
- Catalog quality scoring and the action queue
- Conversation transcripts
Pro Our recommendation
For stores that want to act on what the agent learns, not just deploy it.Runs onYour own server, database and login — shared with no other store.
- Everything in Starter, plus:
- Sized for around 3,000 conversations a month · 5,000 products · 500 pages
- A stronger model on hard questions — escalates when the fast model isn’t confident
- Six months of knowledge-gap history, not just the last week
- Catalog quality scoring — per product, with the specific fix
- Action queue — ranked by the cart value each gap is costing you
- Attribution — question to clicks to cart to checkout, per session
- Coverage over time — is the catalog getting better at answering
- Conversation transcripts, personal details masked
- Priority support
Scale
For stores past a few thousand conversations a month. Pro’s console, with more room.Runs onYour own server, database and login — shared with no other store.
- Everything in Pro
- Sized for around 10,000 conversations a month · 25,000 products · 2,000 pages
- Priority support with a named contact
Enterprise
For deep catalogues, heavy traffic, or a procurement process that needs answering.Runs onYour own server, database and login — shared with no other store.
- Everything in Scale
- Conversations, catalogue size and crawl depth agreed with you at onboarding
- A larger server when your traffic needs it, moved by our team
- Catalog integration built for your stack — feed, API or custom mapping
- Dedicated support and priority onboarding
- Support through your security review — data flows, DPA, sub-processors
- Invoicing and paperwork to suit your procurement — still month to month, like every plan
Find your plan.
Bigger than Starter, smaller than Pro? Put your own numbers in and see which published plan covers them. There is no in-between price: it shows the smallest plan whose limits fit, and we confirm that with you on the call. Every plan still runs on a server for your store alone, which is what sets the $99 floor.
On volume: we don’t bill per conversation, so a busy month never produces a bill you didn’t expect. Plans are sized to your traffic and we agree it with you up front — we won’t switch the agent off mid-month or bill you for overage afterwards. Beyond Scale, or when procurement needs answering, Enterprise is the right shape.
Your store gets its own server. Even on Starter.
Running every store out of one shared database would be cheaper. We don’t. Every plan launches a server for one store — its own database on its own disk, its own login, its own certificate — and that machine runs every hour of the month whether or not a shopper asks it anything. That is what the price buys. The plans differ in what you get on top, not in where your conversations live.
Your own server, database and login
Your shoppers’ conversations are kept in your own database, on your own server, behind your own login — not in one big database shared with other stores. Our central reporting receives only counts and costs.
Encrypted on the way, masked in your console
Traffic between your shoppers and your agent is encrypted. In transcripts, shoppers’ personal details are masked until you choose to reveal one, and every reveal is logged. The central platform rejects anything that contains message content.
Products, not errors, when the AI is down
If the AI behind the agent has an outage, shoppers still see matching products from your catalog and a short note that the assistant is temporarily unavailable — not an error page, and never a made-up answer.
A bigger server on request
Every store starts on the same size of server. If your traffic outgrows it, our team moves you to a bigger one — a short restart, the same software, your data where it was.
What we don’t claim: an uptime figure, a certification, or that nobody at HOLLOU can reach your server — our engineers can, because they run it. Each store is one machine; if it fails, the agent is offline until we restart it. We’d rather you read that here than find it in a questionnaire.
Most assistants fail quietly.
A shopper asks something your catalog can’t support, gets a vague answer, and leaves. You never find out. That happens on every assistant — the difference is whether anyone tells you.
A bolt-on chat widget can’t produce this, because it doesn’t know your catalog well enough to grade it. We index your products and your pages first, which is what makes the answers grounded — and the same index is what lets us tell you where the gaps are. The cost figures are modelled from your own data and labelled as modelled; we don’t compare you to an industry benchmark we don’t have.
For POS vendors, ISOs and agencies.
If you already have merchants, this is a different product — not a discount on the ones above.
Your brand
Shoppers never see our name — not in the widget, not in the transcripts. You resell it as yours, at your price.Provisioned for you
Send us the merchant; we create it and hand back the key and the embed snippet. Each merchant runs on its own server with its own catalog and its own data.Your relationship
You own billing, support and the merchant. We’re infrastructure you put your name on, and we don’t sell to your base.Partner pricing is quoted, because it depends on how many merchants you onboard and how fast. Tell us about your portfolio.
The things people actually ask.
How is pricing structured?
One monthly price per plan — Starter $99, Pro $249, Scale $549, Enterprise quoted. Monthly, with no annual contract. No per-seat fees, no setup fee, and no per-conversation charge. Plans are sized to your traffic and we agree the monthly volume with you up front.
Can we pay annually?
Not yet. Plans are monthly for now — we’d rather you could leave than be locked in before we’ve proved it. We’ll revisit annual billing once the plans have settled.
What’s actually different between Starter and Pro?
Starter shows you that gaps exist — the questions your catalog couldn’t answer over the last week. Pro tells you what to do about them: which product, what’s missing from its record, what it’s costing you, and whether your coverage is improving as you fix things — and answers hard questions with a stronger model, where Starter stays on the fast one. Starter is the cheapest way to find out whether the gap list is worth acting on. Scale is Pro with more room — the same console, sized for around ten thousand conversations and a 25,000-product catalogue, with a named contact.
Are the catalogue sizes hard limits?
No. They describe what each plan is sized for, not a cap the software enforces. We look at your catalogue before we provision anything, and if it’s larger than the plan assumes we’ll tell you then rather than after you’ve paid. Nothing switches off if you add products. Crawl depth is set per merchant at onboarding, so if your site is deeper than the plan assumes, that’s a conversation not a wall. The same goes for conversations: we don’t meter them today, we agree the volume with you, and we tell you before anything changes.
What counts as a conversation?
One shopper working through a question with the agent, however many messages that takes. We don’t bill per conversation, so this only matters when we’re sizing a plan with you. Sessions where the shopper never asked anything don’t count.
What happens if we outgrow the plan?
We tell you, and we agree the next step together. We won’t switch the agent off mid-month and we won’t bill you for the overage retrospectively. Losing a shopper to a billing limit is a bad outcome for both of us.
Is there a free trial?
Not a self-serve one. Book a demo instead: we index a slice of your catalog and show you the agent answering real questions about your actual products before you commit — which tells you more than a trial would, and takes about a day.
What if it can’t answer well from our site?
Then don’t buy it. Answer quality depends almost entirely on how much real content your site carries — a thin site produces a thin agent, and we’d rather show you that up front than have you find out in month two. That is also exactly what the knowledge-gap report measures, so you’ll keep seeing it after you buy.
Do you do competitive pricing intelligence?
No, and we won’t estimate it. Knowing what a rival charges needs an external price feed we don’t have, and nothing in your own catalog implies it. You’d reprice against those numbers, so inventing them would be worse than showing nothing.
Do you charge per seat?
No. Your whole team can use the console. There’s no setup fee either, and no long-term contract on any plan.
Which platforms are supported?
Any website, through one script tag. Shopify has a direct catalog sync. WooCommerce, Magento and custom stacks work through our product API, with a mapping step we do with you during onboarding — see Implementation for the honest version of that. Running Shopify? See HOLLOU for Shopify stores.
Can it invent things about our products?
It is built to answer only from your catalog and your pages, not from the model’s general knowledge of your industry. When nothing in your content matches the question well enough, the turn is flagged as thin — and on Pro and above that turn also routes to the stronger model — and the model is instructed to say it doesn’t know rather than fill the gap. There is no separate fact-checker on the way out, and we don’t claim one. When it does say it doesn’t know, it can offer a link to your contact page if you’ve set one, and that gap lands in your report.
Can we cancel?
Yes, and removing the agent is one line in your template that takes effect on the next page load. No long-term contract on any plan.
Is our data secure, and are you certified?
Your agent runs on its own server — its own database, its own disk, its own login, its own certificate. Traffic is encrypted in transit, and your customer data is never used to train external models. Any shopper’s data can be erased on request; personal details are masked in transcripts by default and every reveal is logged. Your server runs on AWS, in the region chosen for your store: Sydney (ap-southeast-2) or N. Virginia (us-east-1). Our sub-processors are AWS, Anthropic, NVIDIA and Microsoft (Azure OpenAI Service); replies come from Anthropic or Azure OpenAI, whichever is primary for your instance, with the other as backup. We never take a payment and never ask for card details — checkout stays on your platform — so we sit outside PCI-DSS cardholder-data scope. If a shopper types a card number into the chat anyway, it is masked in your console. We are working toward SOC 2 Type II and do not hold a report today. If your procurement needs a current report, we’re not the right fit yet — and we’d rather you heard that from us on day one.
See what your catalog can’t answer yet.
Send us your store URL. We’ll index a slice of your catalog, show you the agent answering real questions about your actual products — and show you the questions it couldn’t answer, before you commit to anything.
No catalog credentials needed for the first preview
Prefer to talk it through? WhatsApp us or write to contact@hollou.ai.