Lovable raised $400 million in August 2026 at a $13.3 billion valuation. It runs roughly $500 million in annualised revenue and hosts 60 million projects. It is one of the fastest growing software companies on earth. Source: https://techcrunch.com/2026/08/12/lovable-confirms-new-13-3b-valuation-raises-another-400m/
The Firecrawl team built an open version of the same idea and put it on GitHub. It is called Open Lovable. It has 28.4K stars and 5.4K forks, and it ships under the MIT license.
One thing to be straight about before you read further. Firecrawl and Lovable are partners, not rivals. Firecrawl sells the web scraping that powers a lot of what Lovable does, and Open Lovable is Firecrawl showing off its own API. The repo README even points you to Lovable at the end. Nobody is being attacked here. The license is simply generous, and that is the opportunity.
What it actually does
You paste a website URL into a box. Four things happen.
One. Firecrawl goes to that URL and reads the page. Layout, copy, colours, structure, all pulled back as clean data.
Two. An AI model rewrites what it read as a fresh React app. The model is yours. You plug in your own key.
Three. The result runs in a temporary sandbox so you get a live preview in the browser straight away.
Four. You chat at it to change things. Make the hero darker. Swap the pricing section. It edits the code and re renders.
That is the whole product. It is a website to React converter with a chat editor on top.
Getting it running, about 15 minutes
Step one. Clone it.
cd open-lovable
Step two. Install. Any of the three works.
pnpm install
Step three. Create a file called .env.local in the project root and paste your keys into it. The next section lists exactly which ones.
Step four. Start it.
pnpm dev
Step five. Open http://localhost:3000 in your browser.
That is it. If you have ever run a Next.js project before, nothing here will surprise you.
The keys you need, and where to get each one
Required, no way around it:
FIRECRAWL_API_KEY. This is the scraper. Sign up at https://firecrawl.dev and grab a key from your dashboard. Docs live at https://docs.firecrawl.dev
One model provider, pick whichever you already pay for:
GEMINI_API_KEY from https://aistudio.google.com/app/apikey
ANTHROPIC_API_KEY from https://console.anthropic.com
OPENAI_API_KEY from https://platform.openai.com
GROQ_API_KEY from https://console.groq.com
One sandbox provider, this is where the generated code actually runs:
Vercel is the default. Run vercel link then vercel env pull in your project folder and it fills in VERCEL_OIDC_TOKEN for you. Docs at https://vercel.com/docs
E2B is the alternative. Set SANDBOX_PROVIDER=e2b and get a key from https://e2b.dev
Optional, only if you want faster edits:
MORPH_API_KEY from https://morphllm.com/dashboard
Every one of those URLs comes straight out of the project's own README, so they are the ones the maintainers themselves point at.
What the MIT license actually allows
MIT is about as permissive as software licensing gets. You can copy it, change it, host it, rename it, and sell access to it. No revenue share. No permission to ask for. No lawyer to call.
The full text is here: https://github.com/firecrawl/open-lovable/blob/main/LICENSE
And if you want the plain reading of the license itself: https://opensource.org/license/mit
There is exactly one obligation. The license says the copyright notice and the permission notice have to be included in all copies. People hear that and imagine something heavy. Here is the entire notice you are required to keep:
Copyright (c) 2024
That is the whole thing. No company name, no author name, just that line plus the standard permission paragraph, sitting in a LICENSE file in your repo. Keep it there and you are compliant.
So when people say you can resell this, they are not exaggerating. They are describing what the license plainly says.
What the repo does not give you
This is where most of the resell it as a SaaS advice quietly stops, so here is the honest version.
The repo is the builder. It is not the business.
There is no authentication. Anyone who reaches the URL can use it.
There is no user database. Nothing is saved between sessions.
There is no billing.
There is no deployment for the apps it produces. Previews run in a sandbox and then they are gone.
And every single generation spends your API credits, not the visitor's.
The three things to add before a client pays you
One. Put a login in front of it. Supabase at https://supabase.com and Clerk at https://clerk.com both drop into a Next.js app in an afternoon. Free tiers on both are generous enough to launch on.
Two. Attach billing. Stripe at https://stripe.com is the obvious answer, and their Next.js guides are good.
Three. Meter the usage. This is the one people skip and regret. Cap how many generations a user gets per day or per plan, because one enthusiastic visitor can empty your provider account overnight while you sleep.
None of that is exotic. It is a weekend or two for someone who has shipped before. But it is real work, and it is the part you are genuinely selling. The builder is the commodity now. The wrapper is the product.
Who this is actually for
If you run an agency, this is the strongest case by far. Host it internally, use it to spin up client sites in an afternoon, hand over the repository at the end. Your client never needs to know what produced the first draft, and you never pay a seat fee to a vendor again.
If you want to launch a builder for a narrow niche, this is a serious head start. A version tuned only for restaurant sites, or only for property listings, is a far easier thing to sell than a general purpose builder going head to head with a company worth thirteen billion dollars.
If you just want to build your own sites for free, clone it, add your keys, and never think about the licensing question at all.
The one mistake to avoid
Add the login before you put it on a public URL. Not after. The repo ships with no authentication at all, and the moment it is reachable, every generation is billed to your account by whoever finds it. Lock it down first, then worry about pricing.
Every link in one place
MIT explained: https://opensource.org/license/mit
Firecrawl, required key: https://firecrawl.dev
Firecrawl docs: https://docs.firecrawl.dev
Google Gemini key: https://aistudio.google.com/app/apikey
Anthropic key: https://console.anthropic.com
OpenAI key: https://platform.openai.com
Groq key: https://console.groq.com
E2B sandbox: https://e2b.dev
Vercel docs: https://vercel.com/docs
Morph, optional faster edits: https://morphllm.com/dashboard
Supabase auth: https://supabase.com
Clerk auth: https://clerk.com
Stripe billing: https://stripe.com
Lovable valuation source: https://techcrunch.com/2026/08/12/lovable-confirms-new-13-3b-valuation-raises-another-400m/
You are reading The AI Leverage, where we take one tool a week apart until you know exactly what it does, what it costs, and what it cannot do yet, so you get sharper every time you open it.
by bestapps.ai on instagram