We get some version of this question in almost every onboarding call, usually phrased carefully, like the person asking expects a lawyer's answer and is bracing for it. Fair enough — a lot of platforms print "you own your content!" somewhere in the footer and then bury the actual terms three clicks deep in a EULA nobody reads. So here's the direct version, question by question, no footer asterisks.
Do I own the code my agents write?
Yes, fully, from the first commit. The moment a build finishes — website, app, game, whatever — the source is yours under a permissive license with no royalty, no revenue share, no "as long as you keep paying us" clause. We don't hold a claim on it because we didn't write it in any sense that matters legally; you directed it, our agents executed it, and the output is a work product you commissioned. This is the same logic that's applied to any tool-assisted work for decades — nobody thinks Adobe owns a piece of the poster you made in Illustrator.
Can I take my project and leave?
Yes, and we'd rather you knew that before you needed it than found out mid-panic. Every build has a code export sitting right in My Builds — a real repository, not a zip of minified output you can't read. Clone it, push it to your own GitHub, self-host it, hand it to a contractor who's never heard of BuildMidas. It runs. We deliberately avoid the trick some site builders pull where the "export" is technically true but practically useless — a pile of framework-specific markup that only renders correctly inside their runtime. If leaving feels like defeat to a platform, that platform built its retention around exit friction instead of a product worth staying for. We'd rather compete on the second thing.
What license covers the AI-generated code itself?
MIT, applied to your export the moment it's generated. Practically that means: no attribution requirement back to us, no restriction on commercial use, no clause that gets weird if you sell the company. We picked MIT specifically because it's the license builders already understand — if you've ever pulled a package off npm, you've read this license before, which means your lawyer has too and won't need a billable hour to explain it to you.
Who owns the content — copy, images, generated assets?
You do, same as the code, with one nuance worth stating plainly: some AI image and asset generation runs on third-party model providers under the hood, and their output-ownership terms are what govern at that layer (this is standard across the industry — nobody's model provider hands over unconditional IP rights on generated pixels, and any platform telling you otherwise hasn't read their own vendor contract). What we guarantee on our end is that we assert no additional claim beyond what the underlying provider already requires, and we pass along attribution-free usage rights wherever the provider's terms allow it. If you're building something where the provenance of a specific asset matters — a trademark filing, a case that might end up in litigation — that's worth a real legal read of the specific provider's terms, not a blog post's word for it.
What happens to my data if I cancel?
Your builds and their exports aren't deleted the moment a subscription lapses. There's a retention window — long enough that "I forgot to renew during a busy month" doesn't turn into "I lost six months of work" — after which inactive projects are cleaned up to keep storage sane. Export early if you're worried; it takes one click and costs nothing, cancelled plan or not. We don't hold projects hostage as a renewal tactic. That's a short-term trick that makes long-term trust impossible, and trust is the only thing that makes a platform like this work at scale.
Can BuildMidas use my app to train its models?
No, and this one's worth being unambiguous about rather than clever. Your prompts, your code, your data, your users' data — none of it trains anything, ours or a vendor's, without you opting in explicitly somewhere we can point to. The tenant isolation model that keeps one customer's data from ever touching another's applies to model training the same way it applies to database queries: by default, your stuff stays your stuff, walled off, not aggregated into some pool that makes the product smarter for everyone at your expense.
What about the open source dependencies my build pulls in?
Those keep whatever license they shipped with — MIT stays MIT, Apache-2.0 stays Apache-2.0, and if an agent ever reaches for something copyleft like GPL, that's flagged in the build's dependency manifest rather than silently vendored in. We're not going to pretend a package.json full of other people's work becomes ours to relicense just because an agent typed npm install. Read the manifest before you ship something commercial with unusual dependency terms; that's just good practice regardless of who or what wrote the install command.
Is there a catch?
Genuinely, the closest thing to one is the model-provider nuance above on generated assets, and even that isn't a catch so much as an honest disclosure about a layer of the stack we don't fully control. Everything else — the code, the content you write yourself, the data your app collects, the right to walk away with all of it intact — is yours without a hook in it. We think that's the only version of "own what you build" that means anything. The alternative, where ownership is technically true but practically inconvenient enough that nobody exercises it, isn't ownership. It's a marketing claim wearing ownership's clothes, and builders can tell the difference eventually, usually at the worst possible moment.
