You have paid for this system for six years. Could anyone else take it over?

One command, on your own machine, right now

Ancient Code is free and open source. It reads the repository you point it at and prints what it found. No account, no upload, nothing installed permanently, no configuration.

$ npx ancient-code .              # the five measured questions, on your repository
$ npx ancient-code . --report     # the same thing as a page you can forward

That is the same command that produced every number in the four public reports on this site. The sixth question, who holds the keys, is answered by a person and the report says so rather than printing a confident guess.

Needs Node 20 or newer and a full clone, because ages come from git history. In a shallow clone git dates every line to the day it was fetched, so the report says the age was not measured rather than printing a confident zero.

Just the scaffolding question

ancient-fences answers one of the five on its own, and ships inside the first. Use it when that is all you want.

$ npx ancient-fences .            # what is standing in this codebase
$ npx ancient-fences . --check    # and whether those reasons still hold
$ npx ancient-fences . --tasks    # the dead ones, written as work for your agent

Your code never leaves your machine. The only network call in either tool is asking a public tracker whether an issue is still open, and only when you pass --check.

The oldest one in webpack's source, in full: a comment in lib/UseStrictPlugin.js explaining why the code strips a "use strict" it will put back later, with a link to webpack's own issue #1970. That line has not been touched since February 2018. Nobody has had a reason to look at it since, which is exactly why nobody has checked whether the reason still holds.

Both projects are maintained by excellent engineers. That is exactly the point: this is not negligence, it is blindness that nothing on the market corrects.

Findings in tests are counted separately and never listed as work. A comment in a test that links to an issue is usually the regression test for that bug, so a closed issue there is a reason to keep the test, not to delete it. eslint has 399 of those next to 43 in its source, and one number for both would have made it look ten times worse than it is.

The scanner does not edit your code. Knowing that a fence is dead is the scarce part, and every editor already ships something that can do the deleting, so the findings come out as work for whatever agent your team already uses, with the test suite as the safety net.

Dependency creeps up on you one quote at a time

Nobody signs up for lock-in. It arrives over years: a system grows, the people who built it are the only people who understand it, and one day a quote arrives that feels wrong, but you have no way to say why, and no way to leave.

The asymmetry is not about honesty. It is about memory. Your vendor remembers everything they built for you. You remember the invoices. That gap is the whole problem, and it widens every year.

Six questions, five of them measured

Each one is a fact about your repository that a machine can check and anyone can re-check, rather than an opinion about somebody's work. Here they are on a repository you already know.

Nothing above is a sample. It is the output of npx ancient-code . on a full clone of webpack, and it changes when webpack changes. That is the point of measuring rather than judging: the number moves, and you can see which question moved it.

Not an accusation. A memory.

Ancient Code will never tell you a price is unfair. Nobody can measure that honestly, because most of the work is thinking rather than typing, and a three-line fix can legitimately cost forty hours.

What it does is give you back the memory your vendor already has. You stop guessing and start asking the right question.

Your vendor holds the repository

This is the real obstacle, so let us be direct about it. There are three ways in, in order of how well they work:

What Ancient Code is not: it does not monitor individual developers, rank them, count their hours or read their messages. It looks at what was delivered and what state it is in. The unit of measurement is the codebase, never the person.

What people ask first