Download websites that use Module Federation

About Module Federation

Module Federation is not a product anyone buys. A bundler feature that shipped in a major Webpack release, it lets separately built JavaScript applications load code from each other at runtime, so a page can be assembled from parts that different teams deploy on different schedules. Detecting it means recognising the runtime plumbing that arrangement leaves behind, not identifying a vendor with customers.

The reason to care is organisational rather than technical. Splitting a front end this way costs coordination and only repays the cost when several teams need to ship independently into one interface, which usually means a large product organisation, a platform with third party extensions, or a site builder whose output inherits the pattern from a template. Presence is a hint about the shape of the team behind a site, and for vendors selling build tooling, observability, design systems or front end engineering talent, that hint is the qualifying question they normally have to ask on a call.

The distribution asks for caution rather than confidence. As of September 2026 we detect it on 651,319 sites, of which 293,395 are in Russia, while the domain ending we record most often is .com. A single national cluster of that size in an architectural pattern rarely reflects thousands of independent decisions. The likelier explanation is one widely deployed platform or template whose generated bundles carry the feature by default, and anyone using these rows to argue that micro frontends are being adopted in a particular market should test that explanation before publishing the claim.

The same caution applies to the age of the pattern. Architectural fashions in the front end move quickly, and a technique that was being written about everywhere a few seasons ago may be entering production only now, or may already be losing ground to alternatives that solve the same coordination problem inside a single build. A count on its own cannot tell those two stories apart, which is why the direction of travel between snapshots matters more than the number in any one of them.

Detection limits point the same way. The signals are recognisable container globals and chunk loading patterns, both of which a build can rename or obscure, so the count understates deliberate use and may overstate it wherever a generator emits the pattern regardless of whether anything relies on it. No version is recorded, and because the feature belongs to a build tool rather than to a shipped dependency, there is no meaningful release string to record in the first place.

Introduced in 2020 and available under MIT, the pattern is young enough that the useful measurement is direction rather than level. A file with one row per domain lets you separate the templated cluster from the rest by looking at what else appears on those domains, count genuine adoption inside a segment you care about, and compare snapshots to see whether the pattern is spreading beyond the platforms that ship it for free. None of that is reachable from a single domain lookup, which can only tell you that one page happened to load a federated chunk on the day you asked.

With the caveats attached, the column works best as a starting filter. Take the domains, join them to company size, country and the rest of the stack, and the subset that survives is a short list of front ends complicated enough to be worth selling to.