Mantyl.devUse cases · one handover, many seats

Selling a micro SaaS built with AI

You built a product with AI agents, it makes money, and you are selling it. Somewhere in the process a buyer, or their technical advisor, is going to open the codebase and decide how much risk lives in it. Their uncertainty is priced into their offer. Reducing it is the highest-leverage work you can do before listing.

How buyers price what they cannot see

Acquisition due diligence on a small SaaS is short and brutal: does it run, can someone else maintain it, is the revenue attached to code the buyer can actually own. When the answer to any of these is unclear, buyers do not investigate, they discount, because investigating costs more than the deal is worth. AI-built products carry an extra layer of this: the buyer knows most of the code was generated, knows nobody line-reviewed it, and prices that in whether or not it is fair. Every hour their advisor spends failing to get your project running from the README converts directly into a lower number or a walked deal.

Diligence as an artifact, not an interview

A project passportmoves the diligence from interview to artifact. Generated from your repository and agent history, it shows the buyer the architecture and services, setup executed from clean in an isolated sandbox, the decision history behind the product’s shape, and, credibility’s real currency, the honest list of unknowns and unfinished edges. A diligence document that admits weaknesses is believed about strengths; one that claims perfection is re-verified from scratch, at your expense, in deal time.

Then let the buyer check it. Their advisor runs npx mantyl receive against the code they were given and watches the manifest, commit and recorded checks reproduce on their own machine. For the deal itself, a Mantyl Verified run puts a signature on it: the checks re-executed on infrastructure neither side controls, bound to the exact commit being sold. £19 to £49 against a five-figure deal is not a cost, it is the cheapest point of leverage in the whole process.

Timing, and one honest warning

Do the first run months before listing, not the week of. The first passport on any real project surfaces things: an undocumented environment variable, a claim the code no longer supports, tests that quietly stopped covering anything. Those are exactly what a buyer’s advisor would find, so find them first, fix them cheaply, and re-run; the passport you list with should be one whose checks pass because they genuinely pass.

The warning: a passport proves the code runs, reproduces and matches its description. It does not prove revenue, security or product-market fit, and a buyer who reads it as a quality certificate should be corrected, by you, in writing. The mark that keeps its meaning is worth more to your sale than one that oversells; that principle runs through everything on the truth modelpage. If you are on the other side of this table, the buyer’s version of the process is in verifying AI-built software you paid for.

Guides · Docs · PricingWritten against the real product