Imagine this same agent app installed on each. Which one are you holding?
2Pick a task
A real job you'd give the agent. LYBO OS runs the very same one on any device above.
Same signed skill pack — LYBO OS just picks the runtime.
OS providesLYBO OS adds
Every phone and PC now ships its own built-in AI — but each one is different, and none of them is safe to ship a real product on by itself. LYBO OS is one agent you write once: it uses whatever AI the device already has, fills in what that OS is missing, and keeps everything governed. Pick a device and a task on the left, then press Run and watch the same job play out.
1Device chosen
2Task chosen
3Press Run
4Read the result
What just happened
Model router — the live decision
LYBO OS looks at your device and the task, then picks the best "brain" available. Orange = the one it chose.
Capability map — leverage vs augment
Grey = features the OS already gives us (we lean on them). Orange = the governance and skills LYBO OS adds on top.
Agent session — the actual run
Your task goes in; a validated, structured answer comes back — on-device.
Press Run on this device to start the agent…
Guardrails — why it can't hallucinate into trouble
The safety rails that run on every request, so the agent is trustworthy in production.
Schema-first: outputs validated + repaired against a declared JSON schemaConfidence gate: low-confidence → deterministic floor, never a guessGrounded: answers constrained to signed pack knowledge, not open webHITL consent: side-effecting tools (payments, messages) held for approvalGlass-box trace: every route, skip, tool call and token logged with an idEval gate: packs must pass an eval suite before signing/publish
Run a task to see the guardrail trace…
Central nervous system — three tiers, one mesh
The bigger picture: devices stay private and light; heavier reasoning rolls up only when needed.
this device · Tier 1 Edge (TLM/OS-AI)──▶vehicle-node-3 · Tier 2 Node (SLM + local RAG)──▶@cloud.hub · Tier 3 Core (LLM)field-tablet-02kiosk-07
Tier 1 handhelds structure inputs locally (no raw audio upstream) → Tier 2 vehicle/site nodes aggregate, dedupe and answer from regional RAG → Tier 3 core consolidates high-signal summaries into business memory. Heavy compute only when deep reasoning is required — and the hub can never bypass on-device consent.
👋 What you're about to see
Every device now has its own built-in AI — Gemini Nano on Android, Foundation Models on iPhone, Phi Silica on Windows. They're all different, and none is production-ready on its own. LYBO OS is one agent that runs on all of them, using each device's AI and adding the safety and skills it's missing. This demo lets you watch that happen in three clicks:
1Pick a device — the phone or PC the agent is running on.
2Pick a task — a real job, like turning a messy note into clean data.
3Press Run — see which AI it chose, the answer it produced, and a plain-English "what just happened".