What are your repositories actually holding?

$ estimate

Where the numbers come from

The count starts with the lines you entered, then makes three adjustments before anything is priced. All three are visible in the output above rather than buried in a coefficient.

Supporting files are added. When a developer says forty thousand lines, they mean application code. The eligible count also takes configuration, schemas, migrations, templates, infrastructure definitions, SQL and documentation. On a delivered project that is a large share of the tree, none of it is vendored or generated, and all of it counts.

Tests are added for you. Codebases are enriched during preparation, most often by writing the tests the project never had. That happens at no cost to you and it increases the volume you are paid on. This is the one adjustment that surprises people, and it is the reason a real estimate tends to come in above a back of the envelope one.

Vendored code is stripped. Dependencies, generated output, binaries and third-party code are removed first. Nobody pays to license a second copy of a package everyone already has.

Then the part most people miss entirely. Git history is counted separately and it is the larger number by a wide margin. Every commit carries a diff, a message and a position in a sequence, and years of that runs to roughly twenty to thirty times the token volume of the source it produced. The rate on history is lower and the volume is so much higher that history is still where most of the money ends up.

The practical consequence: do not tidy your history before offering a repository. Squashing a branch, rewriting commits, or starting a fresh repo to make things look neat destroys the majority of the value.

Rates last verified 14 September 2026: 100 US dollars per million tokens of source, 40 per million tokens of git history, paid to the code owner rather than split. Partner's published standing rates. They sit in the licensing agreement and can change on 30 days' notice, never retroactively. What a codebase actually earns depends on the sale and on buyer demand, so treat this as arithmetic rather than a quote.

This is arithmetic on published figures, not a quote. Your real per-repo number comes from an actual scan and will differ. Connecting a repository is read-only, costs nothing, and you can withdraw before anything is presented to a buyer.

What actually moves the number

Volume matters less than people expect, and originality matters more. A large repository that is mostly generated scaffolding or a vendored framework is worth far less than its line count suggests. A smaller codebase with years of real commit history, written by people solving a specific business problem, is worth more than its line count suggests.

Language matters too, though not in the direction people assume. Scarcity wins. There is a great deal of public JavaScript in the world and much less public code in the older stacks that quietly run large companies.

And the things that zero it out entirely are ownership and cleanliness. Code your client owns is not yours to license at any volume. A repository with live credentials in its history is a liability first and an asset second.

Worth reading before you get attached to a number: what disqualifies a repo and who owns the code you wrote for a client.