The questions we always get asked.

It is software that gathers the messages arriving from different channels — WhatsApp, email, Instagram, Messenger, phone, your website chat and your app chat — into one shared inbox, and ties them to a single contact and a single conversation. The benefit is organisational rather than technical: whoever answers stops hopping between tools and sees that person’s entire history in one place.

In how it separates audiences. Most platforms assume a company talks to one kind of counterpart; here customers, partners, suppliers or agencies are a first-class dimension, each with its own help center, its own knowledge base and its own AI agent. A partner-only document is not hidden from a customer: for that conversation it does not exist, because the separation is enforced by the database as well as by the application.

It can, as with any generative system, which is why the agent is tied to your own content rather than to the model’s general knowledge. Every reply carries a "Sources" button showing an agent which articles it was built on, and you decide with explicit rules which topics must always reach a person. When a question falls outside the configured perimeter the agent does not improvise: it hands the conversation to a human with the context already assembled.

Not to get started; yes to integrate it deeply. The widget is one line of code on your site and channels are connected from the interface. You only need someone technical to hook up your back office or CRM through the public API, or if you choose to run the platform on your own servers instead of ours.

The main infrastructure is in the European Union, on servers in Germany, and each company sees only its own data: the database itself enforces that, not a check written in the application. For your customers’ data you are the controller and we are the processor, under an agreement compliant with Article 28 GDPR. Nothing from your workspace is used to train artificial intelligence models.

Yes, through Meta’s official connection, not through unauthorised automation on a handset. That means approved message templates, campaigns carrying the customer’s own name, and the contact windows Meta imposes. On the customer’s phone exactly one chat stays live per number, as it does in reality.

The public help center handles 15 languages, each with its own indexable address, and the AI agent replies in whichever language the customer writes in. The agent interface is available in Italian, English and Arabic, with right-to-left support.

Yes. Besides the version we host, the platform installs on your infrastructure with Docker Compose, under a licence with support. In that case the database, the files and the conversations physically stay wherever you decide.

We do not publish a price list, because the price depends on how many people answer, how many audiences you run and how much work the AI does. The model is per active agent seat, plus a metered component for automated replies with a ceiling you set. Half an hour of conversation is enough to get a number.

Almost entirely down to how much you have already written. Connecting the channels and publishing the widget is a matter of hours; what takes time is assembling the content the agent will rely on — and if you already have a help center or an archive of replies, it gets imported rather than rewritten.