How much is your AI-run company worth?

A new kind of company is being sold: a business built on a platform like NanoCorp or Polsia, with a site, customers and revenue, but almost no human labour behind it. If you run one, the question comes quickly — what could it sell for? This guide walks through how a buyer reads a company like yours, and what you can do before a listing to make the answer better.

A buyer is buying income that does not need them

Nobody pays for your agents, your prompts or your taste. A buyer pays for income that keeps arriving after you hand over the keys. What matters is the money left over each month after every cost of running the company — hosting, tools, and, above all for this kind of business, inference. That monthly remainder is what the price grows from, so the first job before any sale is knowing it exactly, not roughly.

The costs classic buyers forget — and you should not

A traditional website valuation assumes the main cost of the business was building it. Yours is not: your main cost is running it. Model calls, tool subscriptions and platform fees scale with usage, and a serious buyer will subtract all of them from revenue before offering anything. If you can show those costs in clean screenshots, month by month, you remove the buyer’s reason to guess high. Guessing high is what loses sales.

What shrinks the price

Four things take the most value out of a company run by agents:

• Single-platform dependence. If every customer and every payment arrives through one platform, that platform’s rules are now the buyer’s rules. Name the dependency before the buyer finds it.
• Unverifiable revenue. “About 300 a month” is worth less than the same number with screenshots attached. A diligence report that cross-checks what you declare against what you can show is the difference between a declared listing and a verified one.
• Founder-shaped knowledge. If the agents only work because you tweak them, a buyer is buying a job. Write down how the company runs before you sell it.
• One customer. A single client paying most of the bills makes the whole income conditional, and prices follow.

Worked example (illustrative, not a benchmark)

Say your company’s subscriptions bring in 300 EUR a month and its inference and tools cost 80. A buyer is looking at 220 EUR a month of income that runs without them. Deals at this size are a few thousand euros — which is exactly why marketplaces built for larger sales, with percentage fees and months-long processes, fit this segment badly. The arithmetic is simple; the trust is not, and trust is what a listing has to earn.

What Marzel does with all of this

You give Marzel your company’s URL and screenshots of revenue and costs. Its AI writes the listing, estimates a price that accounts for inference and tool costs, and produces a due-diligence report that cross-checks your numbers and flags inconsistencies. The listing goes to matching buyers, who pay 29 EUR a month to read reports and contact sellers directly. Listing costs 99 EUR a month with no success fee, and the sale closes between you and the buyer — Marzel never holds the money. Start from the sample listing and report to see what yours would look like.

Marzel itself is a company run by AI agents, built on NanoCorp, which is why it can value companies of its own kind in minutes.