AI in your browser that supports MCP Servers, no network call, no costs, 100% private

Phi-3-mini in the browser on 10 hand-written WGSL kernels at ~80% of WebLLM's speed — no TVM, no compiler, no runtime.

Free, private AI in your browser

Scaffold for browser-local LLM chat apps. Runs Gemma 4 entirely in the browser via MediaPipe + WebGPU — no server, no cloud, no tokens. Mobile-first PWA, Atomic Design, i18n (DE/EN), multi-conversation. Fork it. Own it. Ship it.

Run small LLMs directly in your web browser, no cloud computing needed.

On-device RAG app (WebLLM + Transformers.js + pdf.js). Drop a PDF, ask questions, get cited answers. Runs fully in the browser; data stays local.

Starling - Lightning Fast Autofill Extension

Offline-first AI assistant (browser LLM + RAG) for newcomers & disconnected regions. No API keys, works in airplanemode.

Browser-native LLM chat powered by Transformers.js, WebAssembly and WebGPU.

Terminal-style portfolio website that runs a local LLM in your browser. Type `ask` to chat with an on-device LLM (Gemini Nano / WebLLM via WebGPU) — no backend, no API keys, your messages stay local. Built with React + Vite.

Pushing fused WebGPU transformer kernels to max model size — int4, tiled FFN, Phi-3-mini 3.6B in Chrome

Privacy-first document editor with local in-browser LLM inference via HuggingFace Transformers.js, WebGPU, and Tiptap.