Download websites that use Doxygen

About Doxygen

Nothing is running when we detect Doxygen, and nothing has been for a while. What we match is the residue of a program that finished its work long before the page was requested. Documentation is generated on a developer machine or inside a build pipeline, and the HTML it produces is copied onto a web server that has no idea what made it.

The tool has existed since 1997 and grew up alongside the C++ projects it was written to document. Generated API reference, class hierarchies, call graphs and file listings all carry a recognisable structure, and it is that structure rather than any server behaviour that our fingerprint matches. A hit therefore means one specific thing. Somewhere on this domain, an engineering team published machine generated documentation for a codebase.

We record 18 distinct versions in the field, and 1.8.13 is the build a typical detected page reports. The version stamped into generated output is a fossil rather than a status report, because it records the machine that ran the generator on the day the documentation was built and says nothing about the site today. A page reporting an old build is a page whose documentation has not been regenerated since, which makes the version column a rough indicator of when a project last shipped documentation rather than of what the project uses now.

The tool is published under a GPL v2 licence and our facts name no vendor, which fits the shape of the thing. Software of this kind spreads by being added to a build script, not by being sold, and it surfaces wherever the surrounding culture writes documentation comments in source. Open source project sites, university and laboratory pages, hardware vendors publishing reference material for their libraries, and internal engineering portals that happen to be reachable from outside all land in the same group.

As of October 2026 the fingerprint is visible on 58 sites, and the population it defines is narrow and useful. A domain in that file very likely belongs to an organisation that writes code, publishes an interface or ships a library, which is a sharper filter than an industry classification usually manages. Developer tool vendors looking for projects to integrate with, recruiters mapping engineering organisations, and researchers studying documentation practice across open source all want exactly that filter.

Two limits apply. Documentation often lives on a subdirectory or a docs subdomain, so a per-domain count understates how many distinct projects are represented and lets several unrelated codebases share a single row. Generated pages also outlive the projects that produced them, so a hit is evidence that documentation was once published rather than that anybody still maintains it. The two limits point at the same conclusion. The value of the file is the population it isolates, and asking one address at a time can never assemble that population, because you would need to know the answer before you could ask.