Two ways to build private AI on-device, a members Studio to design and license it, and the enablement toolkits and Academy to help an organisation actually adopt it.
Build on-device AI
Build on the model-agnostic Edge Runtime, extend it across every OS's built-in AI with Lybo OS — then design, simulate and license it all in Edge Studio.
Our own ~2.4 MB on-device agent engine — bounded patterns, durable workflows, local RAG, signed capability packs. Bundle it into your app and ship private AI that works offline.
ExploreThe layer above the platform. One governed agent routes across each OS's built-in AI — Gemini Nano, Apple Foundation Models, Phi Silica — plus a fully-offline sovereign mode.
ExploreYour developer home: simulate conversations, approve actions, inspect traces, and use the visual App Builder. Online, it holds your licence keys, quickstarts and MCP setup.
ExploreLyboAI's model-agnostic on-device agent engine. It ships nomodel of its own — a deterministic tier always answers, then it downloads a small model or bridges any external engine (llama.cpp, Ollama, Core ML). This is the base you build on.
The same runtime, extended: register each platform's built-in AI (Gemini Nano, Apple Foundation Models, Phi Silica) as a governed routing tier, and join an A2A mesh so agents talk device-to-device up to a cloud hub. Not a separate product.
Enable your organisation
The toolkits and courses that turn a licence into real, governed adoption — for Claude, for Microsoft Copilot, and for on-device AI.
The full toolkit to roll out Claude across an organisation — operating model, security, deployment playbook, trainer kit and industry add-on packs.
ExploreRoll out the Microsoft Copilot family — M365 Copilot, Copilot Studio and GitHub Copilot — with tenant governance, department playbooks and add-on packs.
ExploreSelf-paced courses to coach, deploy and govern Claude, Copilot and on-device AI — short lessons, knowledge checks and certificates.
Explore