Desktop experience inside the browser

GadgetOS

Browser-based operating system built to simulate a desktop-like experience entirely within your browser.

Documentation

GadgetOS

An AI-powered web operating system — manage files by meaning, execute in natural language, and build in the browser.

<br />

GadgetOS Desktop

</div>

What is GadgetOS?

GadgetOS is a fully in-browser operating system experience built with Next.js. It features a macOS-inspired frosted glass desktop shell with real apps — a file explorer backed by Convex and S3, an AI terminal powered by GPT-4.1, a multi-tab text editor, an iframe browser with history and bookmarks, a clock, settings, and more.

Features

App	Highlights
File Explorer	Folder tree, file list, image preview, text & semantic (vector) search via Convex
Terminal	`ls`, `cd`, `mkdir`, `touch`, `cat` wired to Convex filesystem — unknown commands answered by GPT-4.1
Text Editor	Multi-tab editor that saves files directly to Convex
Browser	iframe browser with back/forward/reload stack, Convex-backed bookmarks bar and history, server-side proxy to bypass `X-Frame-Options`
Clock	Live analog + digital clock
Settings	Sidebar shell with appearance, profile, and shortcut stubs
About	System info panel
Trash Bin	Soft-delete destination for future file ops
Desktop Shell	Frosted glass navbar, animated floating dock, draggable/resizable windows with dnd-kit

Tech Stack

Framework — Next.js 16 (App Router, React 19)
Database / Realtime — Convex (queries, mutations, vector search)
Storage — AWS S3 (file uploads, presigned URLs)
AI — OpenAI GPT-4.1 (terminal fallback, image descriptions, embeddings)
UI — Tailwind CSS v4, Radix UI, Lucide, Motion
State — Zustand
Drag & Drop — dnd-kit
Analytics — Vercel Analytics (production only)

Getting Started

Prerequisites

Node.js 20+
A Convex account
An OpenAI API key
An AWS S3 bucket

1. Clone & install

git clone https://github.com/Hitesh-s0lanki/gadgetOS.git
cd gadgetOS
npm install

2. Configure environment variables

cp .env.example .env.local

Fill in .env.local:

# OpenAI
OPENAI_API_KEY=sk-proj-...

# AWS S3
GADGETOS_ACCESS_KEY_ID=AKIA...
GADGETOS_SECRET_ACCESS_KEY=your-secret-access-key
GADGETOS_REGION=ap-south-1
GADGETOS_S3_BUCKET=your-s3-bucket-name

# Convex
CONVEX_DEPLOYMENT=dev:your-deployment-name
NEXT_PUBLIC_CONVEX_URL=https://your-deployment-name.convex.cloud

3. Start Convex

npx convex dev

4. Run the dev server

npm run dev

Open http://localhost:3001 in your browser.

Project Structure

gadgetos/
├── convex/              # Convex schema, queries & mutations
│   ├── schema.ts
│   ├── files.ts
│   ├── folders.ts
│   ├── browser.ts
│   └── ...
├── src/
│   ├── app/
│   │   ├── (public)/    # Landing page
│   │   ├── webos/       # Desktop shell & lock screen
│   │   └── api/         # Route handlers (proxy, upload, presign, AI terminal)
│   ├── components/
│   │   └── webos/
│   │       ├── apps/    # All OS apps (terminal, browser, file-explorer, …)
│   │       ├── desktop/ # Navbar, taskbar, desktop screen
│   │       ├── window/  # Draggable window base
│   │       └── ...
│   ├── hooks/webos/     # Per-app Zustand window stores
│   └── lib/             # OpenAI helpers
└── public/              # Static assets & icons

API Routes

Route	Method	Purpose
`/api/proxy`	GET	Server-side web proxy — strips `X-Frame-Options` so sites render in the browser iframe
`/api/upload`	POST	Uploads a file to S3, writes a Convex record
`/api/presign`	GET	Generates a presigned S3 URL for image preview
`/api/openai-terminal`	POST	GPT-4.1 fallback for unknown terminal commands

Deployment

The project is deployed on Vercel. Vercel Analytics is enabled automatically via the VERCEL_ENV environment variable — no additional configuration needed.

Make sure to add all environment variables from .env.example to your Vercel project settings, then deploy:

vercel --prod

Contributing

Fork the repo
Create a feature branch (git checkout -b features/your-feature)
Commit your changes
Open a pull request against the dev branch

Built with ❤️ using Next.js, Convex & OpenAI

</div>

Keep exploring

Related work

All projects

Kya Banana Hai

A smart kitchen assistant that helps you decide what food to cook based on ingredients, time, and mood — powered by AI.

GadgetOS

AI-powered web operating system with a macOS-inspired desktop — manage files by meaning, run natural language terminal commands via GPT-4.1, and build entirely in the browser.

MindBoard Arena

Stateful chess arena where language-model agents reason through the board instead of running engine lines — Human vs AI, Human vs Human, and Agent vs Agent modes with persisted games, move history, and match logs.