htmx

htmx is a small JavaScript library for building interactive pages without a client-side framework. It starts from a narrow complaint about plain HTML: only links and forms can make requests, only clicks and submits trigger them, only GET and POST are available, and every response replaces the whole page.

How it works: you add attributes to ordinary elements. hx-get, hx-post, hx-put, hx-patch and hx-delete send a request, hx-trigger picks the event that fires it, hx-target picks where the response goes, and hx-swap decides how it is inserted. The server answers with HTML fragments, not JSON, so the backend can be any language or framework that renders HTML templates.

What you get: a dependency-free library you load from a script tag or install from npm as htmx.org; hx-boost to turn existing links and forms into AJAX requests that still work without JavaScript; history support through hx-push-url, so the back button restores earlier content; and core extensions for Server-Sent Events, WebSockets, DOM morphing and preloading. For small client-only behavior, the docs suggest pairing it with vanilla JS, Alpine.js or hyperscript.

The limits come from the model itself. The htmx essays say hypermedia fits text-and-image UIs, CRUD forms and screens that update in well-defined blocks, but not UIs with many interdependent parts that change without the server (a spreadsheet), apps that must work offline, or interfaces that update constantly, like games tracking mouse movement. Every interaction is a round trip, and escaping user content against XSS stays your job.