Starter

AI chat app with streaming responses

A ChatGPT-style chat UI that streams tokens from your own server route, so your API key stays secret.

The promptPaste into Cursor, Claude Code, Codex, Lovable, Bolt, Replit or v0

Build a chat app with streaming responses from [OpenAI / Anthropic / another LLM API].

- Server route that holds the API key (server-only env var) and streams the response.
- Client UI: message list, input with Enter-to-send, Stop button, auto-scroll, markdown rendering with code blocks and copy buttons.
- Keep conversation history in the browser; send only the last N messages to control cost.
- Per-IP or per-user rate limiting and a max input length on the server route.
- Clear error messages for 401 (bad key), 429 (rate limit/quota), and network failures.
Use the provider's official SDK and current docs for model names.

Tips

Never call the LLM API directly from the browser with your key.

Works with: Cursor, Claude Code, Codex, Replit, Lovable, Bolt, v0. Last reviewed 2026-11-13.