Download websites that use Kestrel

About Kestrel

Kestrel sits underneath modern applications on one specific vendor stack, and its presence is one of the clearer organisational tells in the index. As of September 2026 we detect it on 3,719,748 sites, and those sites belong to a different web from the one most technology surveys describe.

Applications on this stack are disproportionately internal tools, business-to-business portals, government services and line-of-business systems that happen to be reachable over the public web. They are not built to attract traffic, they rarely appear in popularity rankings, and datasets assembled by crawling popular sites miss them almost entirely. A measurement that starts from the whole domain space rather than from a traffic list is the only way they show up at all.

The population is also, on average, more institutional than the web as a whole. Organisations that standardise on a single vendor platform tend to have internal development teams, compliance requirements and budgets, which changes what selling into them looks like. Volume is low and value per row is high, which is precisely the situation where a complete list beats a large one.

Where it appears matters more here than elsewhere. United States accounts for 3,639,875 of detections, and unlike consumer platforms this concentration is closer to genuine, since organisations of this type more often host domestically for regulatory reasons. The signal is still about hosting rather than headquarters, but the two align more often in this population than in others.

The .com zone leads, though organisational and country-code zones appear at higher rates here than across the web generally, and those rows are frequently the most interesting ones.

The webatla export carries every domain we detect it on with the framework above it, the proxy in front of it, the hosting country and the rest of the identified stack, as one file rebuilt daily.

There is a structural reason this population is worth measuring separately rather than folding into general server statistics. Applications on this stack are frequently reached through a proxy that terminates the public connection, so the component only becomes visible when something in the response path passes its signature through. Our figures are therefore a floor, and the true footprint inside organisations is larger than what the public web reveals.

That understatement runs in a useful direction for prospecting. The sites where the component is visible are disproportionately those with simpler public infrastructure, which tend to be the smaller and more approachable end of an otherwise enterprise-heavy population. The bias is real and worth knowing, and it happens to point toward the more reachable half of the market.

Microsoft publishes it under the MIT license as part of the wider platform, and it has been shipping since 2016, written in C#.