Deploying it on a VPS in 2 minutes
The Docker command is `docker run -d -p 3000:3000 yidadaa/chatgpt-next-web`. I deployed it on my HK server with nginx reverse proxy and a custom domain. The first-run screen asks for an API key. I entered my DeepSeek API key and had a working ChatGPT-style interface. Adding more providers is just adding more API keys in the settings panel. NextChat supports OpenAI, Anthropic, DeepSeek, Google Gemini, Moonshot (Kimi), ByteDance (Doubao), Alibaba (Qwen), Zhipu (GLM), and any OpenAI-compatible endpoint.
The multi-model workflow that saves time
My typical session: start with DeepSeek-V4 for a coding question, switch to Claude for a more nuanced code review, then switch to Kimi-K2.6 for Chinese-English translation work. All in the same chat interface, same conversation history, same keyboard shortcuts. Without NextChat I would have 3 browser tabs open: chat.deepseek.com, claude.ai, and kimi.moonshot.cn. Each with different UI conventions, different keyboard shortcuts, different conversation management. The friction reduction is the main value.
Privacy and data: your keys, your data
NextChat stores API keys and conversation history in the browser's localStorage by default. Nothing goes to a NextChat server. If you deploy it on your own server, data stays on that machine. This matters because the official ChatGPT and Claude interfaces train on your conversations by default (you can opt out, but it is buried in settings). With NextChat, you are calling the API directly with your own key, which means the conversation data is governed by the API provider's privacy policy, not their consumer app policy. API conversations are not used for training on OpenAI and Anthropic by default.
What the official apps do better
NextChat is purely a chat interface. No file upload analysis, no image generation, no voice mode, no plugins. The official ChatGPT app has all of these. If you use DALL-E for images, ChatGPT's built-in integration is smoother than switching to a separate tool. The Claude UI has a better artifact viewer for code output. And the mobile experience for NextChat is just a PWA, no native app with push notifications or offline mode.
When NextChat is the right choice
Use NextChat if you: use 3+ LLM providers and want one interface, need to deploy a private chat UI for a team, want to avoid training-data-opt-in confusion, or have a limited budget and want to use the cheapest API per task (DeepSeek for coding, Claude for reasoning, Gemini for search). Do not use NextChat if you: rely on file uploads, image generation, or voice mode heavily, or if you only use one provider and their official app is good enough.
