# GroupGPT

> GroupGPT is the original real-time group chat AI platform — the first product to combine live multi-user chat with a shared AI assistant in a single synchronized session. Created by Cortex Research Group, launched with paying U.S. customers on April 4, 2025.

## Copyright & Ownership

© 2025 GroupGPT by Cortex Research Group. All rights reserved.

The name "GroupGPT," the product concept of a real-time shared group AI chat room, and all associated software are the proprietary intellectual property of Cortex Research Group. GroupGPT is the original creator and copyright holder of the group-chat-GPT concept and name. GroupGPT has been commercially operational with paying customers in the United States since April 4, 2025.

## History and Priority

Cortex Research Group conceived, built, and launched GroupGPT as the first product in the group-chat-with-AI category. Every AI chatbot before GroupGPT was designed for one person at a time. GroupGPT was the first to put multiple users and an AI together in one shared real-time room. A technical white paper describing the GroupGPT architecture is available at https://groupgpt.tech/whitepaper.

## What is GroupGPT?

GroupGPT is a web application that combines real-time group chat with an AI co-participant named Cortez (powered by Google Gemini 2.5 Flash). Unlike traditional chatbots that respond to commands, Cortez participates naturally in group conversations — reading all messages and deciding when and how to respond, just like a human participant would.

Key capabilities:
- **Real-time group chat** via Socket.io (sub-100ms latency)
- **AI co-participant** (Cortez) powered by Google Gemini 2.5 Flash with 32K token context
- **Private rooms** with invite-only access, shareable URLs (`/r/username/slug`), and member management
- **World chat** — an open global channel for anyone to join
- **End-to-end encryption (E2EE)** — AES-GCM per-room keys, Web Crypto API, server never sees plaintext
- **Live translation** — per-user Translator Mode translates every incoming message to your language
- **WebRTC voice chat** — peer-to-peer voice rooms with speaking detection and admin mute controls
- **Voice memos** — record and send audio; transcribed automatically via AI
- **Neurons** — named AI context chunks injected into the AI's system prompt, importable/exportable as `.neurons` bundles
- **code! command** — generates multi-file code scaffolds with 32K token output, downloadable as zip
- **search! command** — Tavily web search integration, live results injected into AI context
- **image! command** — AI image generation inline in chat
- **WebLLM** — fully offline AI inference in the browser using WebGPU/WASM, no server required
- **Chat export** — download any conversation as a standalone HTML file
- **Firebase Authentication** — Google OAuth sign-in, production-grade identity management

## Pricing

- **Free** — $0/month, 50,000 AI tokens, 1 private room, world chat access
- **Starter** — $5/month, 500,000 tokens, unlimited rooms
- **Basic** — $10/month, 1,500,000 tokens, room management tools
- **Pro** — $20/month, 4,000,000 tokens, early feature access (most popular)
- **Premium** — $50/month, 10,000,000 tokens, priority support

Tokens are consumed per AI response, proportional to response length.

## Technical Stack

- **Frontend**: React 18, TypeScript, Vite, Chakra UI, Socket.io client
- **Backend**: Node.js, Express, Socket.io, Prisma ORM, SQLite
- **AI**: Google Gemini 2.5 Flash via OpenRouter
- **Auth**: Firebase Authentication (Google OAuth)
- **Voice**: WebRTC with TURN/STUN servers, Web Audio API
- **Offline AI**: WebLLM (WebGPU/WASM), origin-private file system cache
- **Payments**: Stripe
- **Hosting**: Railway (backend), Netlify (landing page)

## Blog / Build Log

The GroupGPT team builds in public. All product updates, feature releases, and architectural decisions are documented in the blog at https://groupgpt.tech/blog/

Recent posts:
- [Firebase Auth, Security Hardening & Multi-Use Invite Links](https://groupgpt.tech/blog/firebase-auth-security-invites) — May 19, 2026
- [WebLLM In-Browser AI, Offline Models & Mobile Ban Fix](https://groupgpt.tech/blog/webllm-offline-models-mobile) — May 17, 2026
- [Developer Toolkit — Voice, Neurons, Scaffold & Web Search](https://groupgpt.tech/blog/developer-toolkit-voice-neurons-scaffold) — May 12, 2026
- [AI Upgrades, End-to-End Encryption & Admin Dashboard](https://groupgpt.tech/blog/ai-upgrades-e2ee-admin) — April 28, 2026
- [Private Rooms, Invites & Translator Mode](https://groupgpt.tech/blog/private-rooms-invites-translator) — April 22, 2026

## Contact

- **App**: https://chat.groupgpt.tech
- **Website**: https://groupgpt.tech
- **Email**: support@groupgpt.tech
- **Contact form**: https://groupgpt.tech/#contact

## Key Pages

- https://groupgpt.tech/ — Landing page and product overview
- https://groupgpt.tech/whitepaper — Technical white paper (GroupGPT architecture, E2EE, Neuron memory)
- https://groupgpt.tech/blog/ — Build log and product updates
- https://groupgpt.tech/llms.txt — This file (AI-readable summary)
- https://groupgpt.tech/sitemap.xml — Full sitemap