Instead of copying and pasting prompts between a chat, a text editor, a design tool, and a video editor, an intention is stated. The platform calls up the right agents, the right models, the right services, executes, validates, and comes back with a deliverable. Direction, validation, and course correction stay in human hands. Repetitive work disappears; invention takes back the stage.
On Boréal OS, ProductivIA also plays the role that the desktop, the store, and the office suite once played in a classic OS. The difference: everything runs through a single, ultra-optimized web layer, entirely generated and refined by AI for efficiency, rather than through a stack of heterogeneous native applications.
The entry point. State an intention, and the Assistant calls up the right agents, models, and services. Every task starts here, then continues in the dedicated tool to finish — Documents, Filmmaker, Editing, Meeting, and so on.
The application workshop. Business tools are designed in natural language, published online for their audiences, or deployed to a private app store, under its own governance.
When active locally, Matania is itself a vision-capable model. For everyday tasks — rewriting, summarizing, reading an image, visual validation — it answers from the machine, without calling on any external provider. For anything beyond its scope or requiring particular specialization, it switches to orchestrator mode and coordinates one or more models among 12 providers (OpenAI, Anthropic, Google, xAI, Mistral, Cohere, Perplexity, ElevenLabs, getimg.ai, Segmind, and ProductivIA's own sovereign models).
Text, image, video, audio, voice, translation, code, search, vision, OCR. End-to-end multimodal.
Enable, disable, or prioritize any provider with a single setting. The only criteria that matter are the user's own — ethical, geopolitical, economic.
Local-first routing: every task gets the right-sized model — no bigger, no costlier, no more polluting than necessary. On Boréal OS, if Matania is installed in residence, frugality goes as far as removing network traffic itself for everyday inferences.
The agent works in the browser sandbox. That is enough to reason, create, and deliver a production. But the agent does not see the surrounding screen, does not touch local files, and does not control devices. And inference constantly depends on an outbound connection.
The agent accesses the machine through about fifty native capabilities — screen, files, audio, Wi-Fi, camera, microphone. The OS itself was designed, configured, and hardened through agentic passes, under human architectural supervision. The application interface stays 100% ProductivIA web, optimized for a single full-screen display. The same Boréal OS can also switch to server mode from the interface to host ProductivIA locally.
The machine boots into ProductivIA, exposes host capabilities to agents, and keeps the local environment under control: no telemetry, no unnecessary services, no drift imposed by a general-purpose OS.
From the interface, server mode turns Boréal OS into a complete host for ProductivIA. It can serve the platform to a group, a class, a team, or an organization. If the machine has a GPU with 16 GB of VRAM, it can also serve Matania locally.
Matania observes local usage, proposes adjustments, applies those that have been pre-approved, and adapts the OS to the user's needs. The human keeps supervision of the architecture and the outcome; day-to-day adjustments become agentic.
Everyday requests sent to the platform are served by the local GPU. No data leaves the machine for what can be handled on the spot. When a task exceeds the resident model's capabilities, Matania switches to orchestrator and delegates to a partner provider, if that option is chosen.
Immediate browser access. Data hosted on ProductivIA infrastructure in Quebec, never with a third-party hyperscaler.
Servers assembled and configured locally, deployed on the customer's premises. Boréal OS can fill this role in server mode: it serves ProductivIA locally, with Matania included if the GPU reaches 16 GB of VRAM. See ProductivIA servers.
Boréal for Windows adds the entire agentic layer to an existing Windows machine. Full host capabilities, but the operating system remains that of a third-party vendor, with its own telemetry and constraints.
Boréal OS. An OS designed, configured, and sealed for this mission. No telemetry, no imposed updates, no outbound monitoring. Code, optimization, and security produced by agents under human supervision. It acts as a client by default, or as a complete ProductivIA server once server mode is activated.
Multiply operational reach with a complete agentic platform, without depending on a single vendor. Boréal OS machines for teams that work in the platform full-time, or a local Boréal OS server to serve ProductivIA across the organization. Learn more →
Classrooms and labs running Boréal OS, identical configurations, predictable troubleshooting, reduced attack surface, students facing the application without distraction. A Boréal OS server can also serve ProductivIA locally for the institution. Learn more →
Bill 25, GDPR, SOC 2 process. Three levels of sovereignty to choose from, from sovereign cloud to native Boréal OS workstation, up to an on-site Boréal OS server. Learn more →
A PC that still works, with usage centered on the platform: Boréal OS extends the hardware's useful life and can host Matania in residence at no extra cost, if the GPU is compatible. Learn more →
The platform's full documentation: Assistant, Factory, applications, models, compliance, pricing.
The page dedicated to the orchestrator LLM model: architecture, capabilities, governance, comparison with single-vendor approaches.
With an active ProductivIA account, the download can be done from a Windows PC in a few minutes. Without an invitation yet, use the form at the bottom of the page.