Everything we're asked before signing

A conversation is unlimited messages between your agent and a customer over a 24-hour period. It opens with the first message and closes after 24 hours with no further response. Two messages or fifty, it's the same count. If the same customer comes back three days later, it's a new conversation. The messages you send yourself from the interface, the tests you do before publication and the conversations taken over by a human are never counted.

Nothing stops: the agent continues to respond. Conversations beyond the plan are billed at your plan rate, shown on the Rates page, and appear on your monthly bill. An excess limit is set in your customer area, and an alert warns you at 80% of the package. If the excess becomes regular, the interface notifies you of the plan that would cost you less.

No. Subscriptions are monthly, with no minimum duration. You change your plan or cancel from your customer area, in one click; termination takes effect at the end of the current period and the agent remains active until then.

Yes, all amounts displayed are in euros excl. VAT. The applicable VAT is added at the time of payment depending on your country and your intra-community VAT number. Invoices are available in your customer area.

Yes, with no time limit: the 25 conversations included each month are never billed. A payment card is required when you sign up, but nothing is charged while you stay within the allowance. Each additional conversation costs €1: Découverte remains the cheapest plan up to 49 conversations per month, costs the same as Boutique at 50, and Boutique becomes cheaper from 51. The pricing page simulator calculates this for you.

Yes, in both directions and at any time. The change is prorated automatically: you only pay the difference over the remaining days, and your conversation plan is adjusted immediately.

Around ten minutes for a standard store: connection of the store, rereading the rules, testing your latest requests, publication. No technical skills are necessary and no intervention from your agency is required.

No, the agent starts by reading what already exists: your help pages, your general conditions, your product sheets and your returns policy. You then reread a summary of what he understood and you correct in one sentence what is not correct. This is the only essay requested.

This is even the recommended step. Test mode replays your last real requests and shows you the response the agent would have given. Nothing is visible on your store until you publish.

Anytime, with just one button. You can also restrict his hours, limit him to certain pages, or let him respond only when no one is available.

He says it and hands over, rather than improvising. The agent responds using your data: orders, catalog, help pages. Outside of this scope, it forwards the conversation to a human with the full history and a summary of the request.

Answers are constructed from your sources, and each factual statement — order status, price, stock, lead time — comes from a direct reading of your store at the time of the question. Sensitive subjects that you list (disputes, banking data, legal complaints) automatically trigger a transfer to a human.

Yes. Each conversation is searchable, and a correction can be entered in one sentence. It applies immediately to the following conversations: you never correct the same point twice.

It responds in the customer's language, among 32 languages, without additional configuration. Your rules, your tone and your commercial policy remain the same from one language to another. The welcome message and suggestion buttons are automatically translated into six languages (French, English, Spanish, German, Italian and Dutch): visitors see them in their browser’s language.

Shopify, WooCommerce and PrestaShop. Other platforms are being studied: tell us which one you use from the Integrations page or the Contact page, this weighs in the order of priority. For a custom store, the agent can connect to your API or product feed; write to us to check the feasibility before subscribing.

In France. Conversations and customer data are not used to train any model. A GDPR-compliant subcontracting contract is available on request, and you can trigger the deletion of data from your customer area.

The widget code loads asynchronously after the page is displayed and weighs less than 30 KB. It stays out of the critical rendering path, so it does not affect your Core Web Vitals.

Read access, and nothing else. The agent never changes anything in your store: they don't create returns, reserve items, or refund anyone. He checks, he explains, and he hands over when action is necessary.

You export your conversations and statistics in CSV format from your customer area. At the end of the paid period, the agent is deactivated, the widget disappears from your store and the data is deleted within 30 days.