Guide

How to sell software you have built

How to sell software you have built, whether that means the whole product, the IP, or licensing the codebase as training data.

Updated

The short answer

To sell software you own outright, decide what you are selling: the whole product with customers, the intellectual property alone, or a licence to the codebase for AI training. Check ownership of every contributor's code and open-source licences first. Then document it, value it, and sell through an acquirer, an asset sale, a broker or a direct licence.

First: which kind of selling do you mean?

If you mean selling subscriptions or licences of a working app to customers, that is software sales and marketing, and this guide won't help much. It is about selling the software itself as an asset: a product you are exiting, intellectual property you no longer need, or the source code of a closed or pivoted product, which AI labs and data companies now license to train and test coding models.

Three ways to sell the software itself.
OptionWhat transfersTypical buyerWhat they check most
Sell the whole productCode, brand, customers, contracts, often staffA competitor, a strategic acquirer or an investorRevenue, customer contracts, churn, code quality
Sell the IP onlyCode, copyrights, patents and domain, but no customersA company that wants the technology or the nameClean ownership, patents, how reusable the code is
License the codebaseA copy of the code and its history; you usually keep ownershipAI labs and data companiesCommit history, privacy, open-source content, uniqueness

Why AI buyers want old codebases

A model learning to write software benefits from seeing how real products were built and changed over years: the commits, pull requests, reviews and fixes, not only the final files. Public repositories have largely been used already, so private ones are scarcer. MIT Technology Review reported in October 2026 that AI labs, data companies and companies building their own models were buying code and workspace data from shut-down startups, with one founder's codebase reported at about $25,000. That is one data point, not a price guide.

Steps to sell your software

  1. 1

    Prove you own all of it

    Collect the IP assignment or employment clauses for every founder, employee and contractor who wrote code. Code written by a freelancer without an assignment may still belong to the freelancer. Buyers ask for this first.

  2. 2

    Audit what is inside the repository

    List open-source dependencies and copied code with their licences; copyleft licences can impose conditions on what you distribute. Scan for API keys, passwords and customer data in code, configuration and commit history.

  3. 3

    Decide what you are offering

    Whole product, IP only, or a codebase licence. You can sometimes combine them: license the code for training first, then sell the IP, provided the training licence allows it. Decide the order before you talk to anyone.

  4. 4

    Write it up

    Describe the product, languages, size, age, number of contributors, history length, test coverage and documentation. For training buyers, report repository-level statistics; for acquirers, add revenue and customer details under NDA.

  5. 5

    Get a value range

    For acquirers, value follows the business: revenue, growth and strategic fit. For training licences, value follows the code: size, history, activity and uniqueness. DataBroker's app gives a first range for codebases from repository size, history and activity; treat it as a starting point.

  6. 6

    Pick the route

    Acquirers through your network, an M&A adviser or a startup sale marketplace; AI buyers through a broker, a lab's data team or a wind-down service; insolvency through the administrator or court process.

  7. 7

    Sign in the right order

    NDA before sharing code, letter of intent for the main terms, then an asset purchase agreement (for a sale) or a licence agreement (for training). Deliver through a controlled repository transfer and keep a record.

Pre-sale code audit checklist

  • Signed IP assignments for every contributor.
  • A dependency list with licences, and any copyleft code flagged.
  • Secrets removed and rotated, including from old commits.
  • Customer data, logs and test fixtures with real personal data removed.
  • Third-party assets (fonts, images, paid libraries) listed and checked for transfer rights.
  • Any patents, trademarks and domains listed with owner of record.
  • A short history: when the product started, what changed, why it stopped.

What this guide can't tell you

Sources

SourceWhat it supportsLinkChecked
MIT Technology Review, Oct 2026AI buyers of startup codebases; reported $25,000 codebase examplehttps://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/2026-11-07

Questions people ask

Can I sell software I wrote at my old job?

Usually not. Code written for an employer generally belongs to the employer under employment terms. Check your contract before offering it to anyone.

Can I sell my startup's code after it shuts down?

Often, if the company still owns it and the people with authority approve the sale. In insolvency, the administrator or court decides. Check investor and lender rights, and remove customer data first.

Does licensing code for AI training mean I lose it?

Not necessarily. A licence typically grants a copy for stated uses while you keep ownership. An exclusive licence or an outright sale changes that, so read the terms.

Do open-source dependencies stop me selling my code?

Not usually, but they have to be disclosed and their licences respected. Copied code under a copyleft licence can complicate a sale. A licence scan before you pitch avoids surprises.

What makes a codebase more valuable to AI buyers?

Long, real commit history, many contributors, production use, tests and documentation, and code that has never been public. Small or recently created repositories tend to interest buyers less.

Related

Become a data broker

The DataBroker course and tools: find companies with data to sell, price it, license it and close the deal.