chatgptnextweb/nextchat

NextChat is a lightweight, cross-platform AI assistant app, a polished chat client you can run on the web, desktop, or phone that talks to large language models from several providers. Instead of using the official websites or apps from each AI vendor, you bring your own API key from a provider and NextChat gives you a single, fast interface for chatting with them. The README mentions support for Claude, DeepSeek, GPT-4, and Gemini Pro. The way it works is that you deploy NextChat in one of several ways, a one-click deploy on Vercel, a desktop app for Linux, Windows, and MacOS (the desktop client is around 5MB), an iOS app on the App Store, or running it locally yourself, then you provide an API key for whichever AI service you want to use. The app sends your messages to that service and renders the response with Markdown, code highlighting, LaTeX math, and Mermaid diagrams. Privacy is a stated focus: chat history is stored locally in your browser. Long conversations get automatically compressed to save tokens. You would use NextChat if you already have API access to one or more language model providers and want a clean, fast interface across devices instead of running each vendor's site. It has features for creating reusable prompt templates, sharing conversations as images, plugins for tools like web search and calculators, an Artifacts preview window for generated content, realtime chat, and self-hosted model support via RWKV-Runner or LocalAI. There is also an enterprise edition for companies wanting branded, privately-deployed installs with permission control and knowledge-base integration. The app supports many languages and is built in TypeScript.