root@fr:~$ whoami
We bring AI into your company. And it belongs to you.
A French company that guides your adoption and builds your agents: on your data, inside your tools, in France, on models you control. You keep the judgment, the liability and the key.
The rare profile
Knowing, for each task, which model runs, where, and under which law.
Choosing a model on raw power alone answers only half the question. Our craft is elsewhere: for each task, and for each client, choosing the right model, running it in the right place, and making it compliant with the right jurisdiction. It is this agent-by-agent tuning that turns a sovereignty promise into a guarantee you can audit.
The right model per task
Drafting a legal deed does not call for the same engine as an internal summary. We choose knowingly, not by default.
The right place
Sovereign by default, in France, with no US entity in the data path. Frontier by explicit choice, when the task allows it.
The right law
Every agent carries its jurisdiction. You see, in plain view, under which law each answer was produced.
The sovereign infrastructure
root.cloud is the infrastructure we have operated for years.
root.fr does not start from zero. We already operate root.cloud: French and European managed hosting, AFNIC-accredited for .fr, GDPR by default. It is the infrastructure our agents run on directly, and the proof that sovereignty is not a marketing line but infrastructure we operate.
Managed FR and EU hosting
Web, e-commerce, email, domains, VPS. AFNIC-accredited for .fr, GDPR by default.
Per-customer isolation
Your own instance, your own stores, shared with no one.
Data stays in France
CLOUD Act immunity by architecture, not by contractual promise.
Our principles
Seven principles. One thread: you keep control.
Our approach to AI comes down to seven principles. They are not a slogan: they describe what you get, and what we refuse to do.
Augment humans, do not replace them
- The professional keeps the judgment and the liability; the AI does the repetitive work.
- Every answer cites its sources, so you can verify instead of trusting blindly.
- The goal is to make your expertise faster, never to decide in your place.
Sovereignty first
- Your data and decisions stay yours: you own the agents, you do not rent them.
- Sovereign by default: open-weight models, hosted in France, no US entity in the data path.
- Self-hosting and on-premise available when sovereignty is critical.
Production-ready, not prototypes
- Reliability, oversight and graceful degradation from day one.
- We ship systems that hold under real load, not demos that impress and fall over.
- An agent that is not operable and auditable is not shipped.
Measurable value, not buzzwords
- We start from the outcome you get, not from the hype.
- Every agent targets a concrete process with a gain we can quantify.
- If we cannot measure it, we consider that we have not delivered it.
Transparency and explainability
- No black box: citations on every answer and a complete audit log.
- You see, per agent, which model answered, where, and under which law.
- You understand why an answer was produced before you act on it.
Safety and responsibility
- Guardrails, not just capabilities: what the agent must not do is framed.
- Human oversight is mandatory on decisions that carry liability.
- Per-customer isolation and access control so nothing leaks between cases.
Continuous improvement
- We monitor, measure and iterate, process by process.
- Your AI evolves with your business, your models and the regulation.
- Running and tuning is a lasting service, not a one-off delivery we abandon.
Our ethos
Production systems, not slideware.
We deploy an AI layer that runs on your data within the first few weeks. Sovereignty is proven by architecture, value by processes that work. The rest is noise.
Talk to an expertLet us put your AI under your control.
Start with a conversation: your processes, your sovereignty constraints, and the first agent that saves you time this year.