Be independent, choose your local AI future
Talk, code, write —
with an AI that stays with you
SouverainCode is three interfaces (Talk, Dev, Bureau), smart routing across our Mistral models and an independent server in Lévis, Québec. Online by subscription, or entirely in-house with the SouverainBox.
From $25 CAD/month · Free 14-day trial, no card · Cancel anytime
Three interfaces, one account
The same AI, three ways to work. Switch from one to the other in one click; your documents and your Archipel Memory base follow.
Talk to it, ask your questions
A conversation with our Mistral models: questions, summaries, drafts, search within your documents. Talk never writes to your files: it reads, searches and answers.
- Search within your Archipel Memory base
- Preview of a page or video whose link you paste
- Stop button at any time
A coding agent that acts
It does more than answer: it plans, creates files, writes tests and fixes bugs, right in your workspace. It asks for your approval before writing, and you see every step.
- Missions followed step by step, stoppable
- Workspace with a clear quota
- Routing across our models, according to the plan
Documents and spreadsheets, ready to send
Letters, quotes, invoices, reports: templates, an automatic proofreader and a spreadsheet, with export to Word (.docx), text or print. Your documents stay in your workspace.
- Templates for small businesses and professionals
- Automatic French proofreading
- Spreadsheet (.xlsx) and document (.docx)
An agent that reasons, then acts
The AI does more than talk: it plans the task, carries it out step by step, writes tests and fixes what fails, in your workspace. Every write waits for your approval, every step is visible, and a mission stops in one click.
Our smart routing, across our models
With the DEV plan, the agent picks the right model by itself, with no tuning on your side: fast by default, it automatically escalates to a more capable model as soon as a task gets harder. Discovery and Solo work with the Fast model. Our models, all run in-house, never at a cloud giant: ministral for fast answers, devstral for code, plus document search. In the SouverainBox, it is the same routing and the same models, on your premises.
Our independent server, in Québec
Inference runs on our own infrastructure, in Lévis — no American subcontractor, no cloud giant. Your prompts are never sent to a third party, never used to train a model, never sold. Your conversations are accessible only from your account and can be deleted at any time. For total sovereignty, the SouverainBox puts the AI entirely in-house — your data never even leaves your walls. All the details.
Predictable caps, not unlimited
Every plan has a clear monthly request cap and a compute envelope, both shown in your settings at all times. A flat, predictable rate for serious developers and professionals.
Discovery
- ✓ The agent, on the Fast model (ministral), local
- ✓ 50 requests
- ✓ 100 MB workspace, 50 indexed documents
- ✓ No card required
Solo
- ✓ Autonomous agent (plans, codes, tests, fixes)
- ✓ Fast model (ministral), predictable cost
- ✓ Document search (RAG) on your files
- ✓ 3,000 requests / month
- ✓ 2 GB workspace, 10,000 indexed documents
- ✓ 100% local data, never shared
DEV
- ✓ Everything in Solo, plus:
- ✓ Full routing fast → strong → deep (devstral)
- ✓ 5,000 requests / month
- ✓ 5 GB workspace, 100,000 indexed documents
- ✓ Parallel missions
Or have it entirely in-house, once and for all
SouverainBox
- ✓ SouverainCode and our Mistral models installed, the same routing as online
- ✓ The AI entirely in-house, offline if you want
- ✓ 100% of your data within your walls (Law 25)
- ✓ Brand-name hardware, manufacturer's warranty
- ✓ From the Solo workstation to a 100-person team
Secure payment via Stripe. Paid plans are paid at sign-up — no free trial; the free Discovery trial requires no card. Prices in Canadian dollars: no GST or QST is added to the amount shown. Enterprises seeking full sovereignty are directed to the on-premises solution — the SouverainBox, priced by quote.