// the find
ChatGPTNextWeb/NextChat
✨ Light and Fast AI Assistant. Support: Web | iOS | MacOS | Android | Linux | Windows
NextChat is a Next.js web app (plus Tauri desktop wrapper) that gives you a chat UI over any LLM API — OpenAI, Anthropic, Gemini, DeepSeek, and a dozen others. It's aimed at developers and power users who want to self-host a ChatGPT-like interface with their own API keys, either on Vercel or locally.
One-click Vercel deploy with a single env var actually works and gets you running in under two minutes. The multi-provider abstraction (app/client/platforms/) is clean enough that adding a new LLM is maybe 200 lines. Local-first storage — all chat history lives in browser localStorage/IndexedDB, no backend database needed for basic use. MCP support landed recently and the implementation isn't bolted on awkwardly; it's gated behind an env flag so it doesn't bloat default deploys.
The proxy layer (app/api/[provider]/[...path]/route.ts) passes raw API keys through a Next.js route handler, which means anyone who can hit your deployment can use your keys if CODE protection is weak or misconfigured — there's no rate limiting or per-user auth at the app level. Chat sync across devices requires either UpStash (a third-party service) or WebDAV; there's no built-in server-side storage option, which is a real gap for teams. The codebase has grown into a 15+ provider sprawl with significant duplication across the platforms/ files — each provider reimplements streaming and error handling slightly differently, making bugs in one not obviously apply to others. Desktop builds via Tauri add meaningful maintenance surface area and the iOS app source is still closed.