Local AI infrastructure · In development

The web can run AI. On your hardware.

Oryn is a browser and engine initiative designed to let ordinary websites request AI models running on the user's own device.

Explore the runtime Architecture preview · 2026

ORYN / SYSTEM MAPDESIGN ONLINE

APPLICATIONYour website

JavaScript SDK

LOCAL ENGINEOryn runtime

Permissions · Sessions · Streaming

YOUR DEVICELocal models

Compute · Cache · Artifacts

JavaScript SDKChromium / ElectronLocal Python engineUser-controlled compute

ORYN / QUICK FACTS

Local AI infrastructure, in plain terms.

What is Oryn?: A browser-integrated local AI infrastructure initiative for websites that need models running on user hardware.
Who is it for?: Web developers building on-device AI experiences without owning every runtime and model-lifecycle detail.
How does it work?: A website calls a JavaScript SDK, the Oryn runtime coordinates the request, and a compatible model runs on the device.
What is its status?: Oryn is active development and product-architecture work, not a generally available public release.

01 / ARCHITECTURE DIRECTION

One request. Three clear layers.

Oryn explores the boundary between web applications and local compute: a small interface for developers, a visible runtime for users, and models running on the device.

01WEB LAYER

A developer-facing JavaScript SDK

Web applications describe the task they need rather than implementing model runtimes, downloads, and hardware integration themselves.

Requests
Events
Artifacts

02RUNTIME LAYER

A browser-integrated local engine

The Oryn architecture coordinates permissions, model sessions, compatibility, resource state, streaming, and cancellation.

Chromium / Electron
Local Python engine
Session lifecycle

03DEVICE LAYER

Models run on user hardware

Local compute becomes an explicit capability of the browsing environment, with model storage and execution kept visible to the user.

Local models
Model cache
Device compute

02 / RUNTIME RESPONSIBILITIES

The difficult parts belong in the platform.

The architecture is aimed at giving applications a consistent way to work with local models without hiding what is happening from the user.

Permissioned access

Give users a clear boundary between a website and local model execution.

Model compatibility

Match a request with models and formats that fit the available runtime.

Download & cache lifecycle

Design for model acquisition, reusable local storage, and lifecycle visibility.

Resource visibility

Surface device constraints so applications can adapt before starting work.

Streaming & cancellation

Support long-running local jobs without turning the browser into a black box.

Binary artifacts

Move beyond text with outputs such as audio, files, and other generated assets.

03 / INTENDED USE CASES

Different models. One local boundary.

Oryn's design work considers transcription, chat, and music generation—different workloads that need the same clear lifecycle.

WEB REQUESTAudio stream

ORYN RUNTIMESpeech model

LOCAL RESULTLive transcript

DESIGN TARGET / Transcription

A note-taking website requests a compatible local speech model and receives text while audio is processed on the device.

01Permission granted
02Model ready
03Tokens streaming

04 / PRODUCT PRINCIPLES

Local AI should feel like a capability of the web—not a pile of runtime decisions every site has to rebuild.

Visible to the user

Permissions, downloads, active sessions, and resource use should be understandable rather than silently buried inside a page.

Simple for the developer

Applications should express the work they need while the runtime handles model and device integration.

Honest about the boundary

Local model execution does not mean every website is entirely local. Each integration still needs a clear account of its data flow.

DEVELOPMENT STATUS

Architecture in motion.

Oryn is development work and product architecture. This site describes the direction of the system—not a claim that every capability is publicly released today.

Explore OCNO

05 / COMMON QUESTIONS

What Oryn is—and what it isn't yet.

Is Oryn publicly available?

Not as a generally available product today. Oryn is an active infrastructure and product-architecture initiative.

Does Oryn mean every website becomes fully local?

No. The architecture targets local model execution. A website may still use remote services, so each product must describe its own data flow clearly.

What kinds of models is the design considering?

Current design use cases include speech transcription, language model chat, and music generation, with model compatibility treated as a runtime concern.