finds.dev← search

// the find

firecrawl/open-lovable

★ 28,570 · TypeScript · MIT · updated Nov 2025

🔥 Clone and recreate any website as a modern React app in seconds

Open Lovable is Firecrawl's example app for chatting with an LLM to scrape a live website and regenerate it as a working React app inside a cloud sandbox. It's aimed at developers who want a working reference for the 'AI builds an app in a sandbox with live preview' pattern, not a finished product to run as-is.

The generation loop is a real feedback loop, not one-shot: streaming code generation and apply routes are paired with vite error detection, reporting, and auto-restart, so it actually recovers from broken output instead of just dumping code. It's not locked into one vendor — four LLM providers (Gemini, Anthropic, OpenAI, Groq) and two sandbox backends (Vercel Sandbox, E2B) are supported behind config. The scrape-to-app pipeline (screenshot, enhanced scrape, brand style extraction, then code gen) is a genuinely useful reference for anyone building a similar AI-coding-in-sandbox tool.

You need working API keys for Firecrawl plus at least one LLM provider before you can run it at all, and Vercel/E2B sandbox credentials on top of that — there's no free or local path to try it. The README is a setup guide only; there's no explanation of how the sandbox, code-gen streaming, and apply steps actually talk to each other, so you're reading route handlers cold to understand the architecture. A large part of the tree — the Pixi.js hero animations, flame effects, and full marketing landing page under components/app/(home) — is Firecrawl's own site chrome bundled into what's supposed to be a learning example, which bloats the repo for anyone trying to extract just the cloning logic. No tests anywhere in the tree, and as a vendor example app it's not built to be a maintained dependency.

View on GitHub →

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →