A ChatGPT clone that cosplays WeChat and dodges the Great Firewall

What it does

ChatGPT Next is a Next.js frontend for OpenAI’s models that apes the look and feel of WeChat’s chat bubbles. It is pitched squarely at self-hosters, offering Docker, npx, and one-click deploys to Vercel, Zeabur, and Netlify, while warning that domains containing chat, gpt, or ai are likely to be probed and blocked. The app supports modern models including o1, o1-mini, gpt-4o, and gpt-4o-mini on both mobile and desktop.

The interesting bit

The standout feature is API key aliasing: you map short public nicknames like firstkey to real sk-... secrets via an environment variable, so friends can log in without ever handling the actual credential. You can even bake the alias into a shared URL, turning a secret into a disposable guest pass.

Key highlights

• WeChat-style UI with green-and-white chat bubbles, responsive across mobile and PC.
• Supports o1, o1-mini, gpt-4o, and gpt-4o-mini.
• API key aliasing lets you share access via public nicknames rather than raw tokens.
• Built-in caution for restricted regions: the default dev environment skips live API calls to avoid account bans, and the docs warn against suspicious domain names.
• Ships as an npm package, Docker image, or ready-to-fork template for Vercel, Zeabur, and Netlify.

Caveats

• The official demo site is already blocked by the Great Firewall, so you’ll need to self-host or trust a community mirror.
• Local development ships with live API calls disabled by default; you must edit app/api/chat/route.ts to re-enable them.
• The documentation and primary concerns are China-centric, which may leave international developers guessing at some hosting nuances.

Verdict

Worth a look if you want a familiar, self-hosted ChatGPT frontend that can be shared with friends without handing out API keys. Skip it if you need a managed, globally available SaaS with enterprise guardrails—this is a scrappy, BYO-key skin over OpenAI’s API.