Sovereign AI agents, hosted in France, that work for you

An AI system that carries out tasks from end to end — not merely answers — under human supervision.

A chatbot answers; an agent acts: it uses your tools and completes the task.

No, and that is a design choice. Your teams write to the agent from the tools they already use: Microsoft Teams, Slack and email — as they would to a colleague, with no account to create, no software to install and no training to schedule. Your customers reach it on WhatsApp Business, through your website chat, by email or by phone, where they already write: that is decisive for a sole trader or a very small firm, whose customers all use WhatsApp. And you supervise the agent from the web dashboard. These connections rely on open standards, including the MCP protocol, and are included in every plan, at no extra cost and with no limit on how many; only platforms that bill on their own side — WhatsApp Business bills per conversation — are passed through at cost, with no mark-up.

No, and it can be checked. Our subscription prices are fixed. The only variable part is usage — AI tokens, telephony minutes, conversations billed by WhatsApp Business, document reading — and we rebill it at cost, with no mark-up at all: we earn nothing on it, so nothing gives us an interest in seeing it rise. It is visible in your client area, as usage statements are issued, you set your own monthly cap beyond which nothing goes without your agreement, and any unused balance is refunded at the end of the contract.

Yes: hosted in France, isolated resources, audited access. GDPR: governed deployment.

Wherever you decide: in France, in your own country on dedicated servers, or on your premises with the 100% Sovereign package. We apply the law of your country and the requirements of your data protection authority, and we write it into the quote.

Yes: safeguards, human oversight and traceability, with a stricter framework for high-risk uses. For an organisation established outside the European Union, the law of its own country applies; at its request, we put equivalent safeguards in place (human review before any action, a log of the criteria, information for the people concerned).

No: we install and run the agent for you, end to end.

A few weeks depending on the use case, after a 30-minute framing session.

An installation fee, then a running subscription; we prepare a tailored quote based on your needs.

Yes: Microsoft Teams, Slack, WhatsApp Business, email, your calendars and office suites (Microsoft 365, Google Workspace), as well as your business software — CRM, ERP, document management — through secure connectors built on open standards, including the MCP protocol. It also connects to the accounts you own on social networks — Facebook pages, Instagram professional accounts, LinkedIn pages — and to Google Business Profile and Google reviews. Trustpilot cannot be ordered directly: the connection requires your own Trustpilot for Business account with access to the API module, your Business Unit ID, an authorised authentication method, and the matching rights and quotas; it is priced on quote, after a feasibility study, with Trustpilot's own costs entirely at your charge and the integration and maintenance cost quoted separately. These platforms belong to third parties: their interfaces and their rules change without notice, publishing in your name stays under your control — the agent prepares, you decide what goes live — and any access fees they charge are passed on to you at actual cost, with no margin, outside the subscription.

It stops and hands over. The agent never guesses: when information is missing or the request falls outside its scope, it passes the case to the right person along with everything it has already gathered. The preparatory work is not lost, and nobody receives an invented answer.

This is the most common concern, and a legitimate one. Two safeguards: the agent cites its sources with every answer — you can see where the information comes from — and it acts within the limits you set. The agents carry out the tasks you designed and authorised them for; the decisions remain yours. Every action towards the outside world is either approved by you or carried out within a scope you have authorised: lawful in your country, bounded, tested before go-live and logged. No decision with a legal or significant effect is taken by an agent.

That is neither the purpose nor the way our agents work. They take on the repetitive part — re-keying, chasing, sorting, document search — and hand back anything requiring judgement. The decision stays human, by design rather than by commercial caution: that is what the AI Act requires for uses affecting people.

The tier is the monthly volume included in your subscription. You are alerted at 80%. A move to the tier above only happens after two consecutive months over the limit, with 30 days' notice, and you can step back down at any time. An isolated peak never changes your invoice.

It is a prepaid reserve covering variable costs re-invoiced at cost — AI tokens, and cloud where applicable — with no margin at all. At tier S, no amount is requested up front: a first top-up is requested at the first alert at 100% of the included consumption. Elsewhere, the amount shown is only a starting estimate. So that your agent never stops for lack of funds, automatic top-up is on by default: it triggers when 30% of a month's consumption is left, never when the wallet is empty. You set the cap, the trigger and the amount, you can switch it all off, and any unused balance is returned to you at the end of the contract.

They are the one-off start-up fees: configuring the agent, connecting it to your tools and to your dialogue channels (Microsoft Teams, Slack, WhatsApp Business, email, website chat, VoIP telephony), testing and training your teams. This is the initial installation, separate from the monthly subscription.