Download websites that use OpenResty

About OpenResty

OpenResty is usually described as a web server, and that description undersells what it is doing. It takes a well understood proxy and embeds a scripting runtime inside the request path, which turns configuration into programming. Teams reach for it when routing logic has outgrown what a config file can express and they do not want a separate application tier just to make a decision about a header.

As of September 2026 we detect it on 19,835,529 sites. Detections of that scale are surprising for a component most site owners have never heard of, and the explanation is that they did not choose it. Hosting platforms and content delivery layers deploy it underneath their own products, so a single provider decision propagates to every customer at once.

That propagation pattern is what makes the population interesting. Detections cluster by provider rather than by industry or size, so the presence of this component is often a reliable fingerprint for which platform a site is really running on, even when the platform is not otherwise identifiable. It is an infrastructure tell rather than a preference.

Version data supports that reading. We see 67 distinct version strings in production, with 1.31.1.1 the most common. A distribution that concentrated around a small number of releases is the signature of centrally managed deployments rather than individual installations, which is the opposite of what a self-installed server looks like.

United States accounts for 6,853,482 of detections and the .com zone leads, again following hosting geography rather than any market for the software itself.

For anyone selling into the hosting or delivery market, this is one of the more actionable populations in the index, because it identifies infrastructure decisions made by companies rather than by individual site owners. A single provider represents thousands of rows here, and finding the providers is a matter of grouping the population by the other technologies that travel with it.

The webatla export makes that grouping cheap. Every domain we detect it on arrives with the version string, the hosting country, the TLD and the rest of the identified stack, in one file rebuilt daily. The clustering work happens locally, where you can iterate on the grouping until the provider boundaries become obvious.

There is a further consequence of provider-level adoption that affects how you should sample this population. Because one hosting decision produces thousands of detections, counting rows overstates how many independent decisions were made. If you are sizing a market of buyers rather than a market of sites, the meaningful unit is the provider, and recovering it means collapsing the population by the fingerprint cluster rather than treating each domain as a separate account.

The same structure makes the data useful in the opposite direction. When a provider migrates its fleet, the change appears across thousands of domains simultaneously, which is highly visible in a diff between two rebuilds and effectively invisible in any single snapshot. Fleet-level movements are among the most commercially significant events in hosting, and they are only observable if you keep yesterday's copy.

OpenResty Inc. maintains the project under the BSD license and it has been available since 2009. Its scripting layer is written in Lua, which is unusual enough in this part of the stack that the choice itself tells you something about the kind of problem it was built to solve.