copy2llm-widget

The copy2llm button UI — a themeable split button rendered in a Shadow DOM (two-way CSS isolation), wrapping copy2llm-core. Extraction is lazy (runs on click, so it stays correct in SPAs).

Most people want copy2llm-react or the <script> snippet instead. Use this directly for non-React / imperative mounting.

npm i copy2llm-widget

import { mount } from "copy2llm-widget";

const handle = mount({ position: "bottom-right", theme: "auto" });
// later:
handle.destroy();

function mount(options?: WidgetOptions, target?: Element): { destroy(): void };

Actions

Split button: Copy as Markdown (primary) · View as Markdown (overlay) · Open in ChatGPT / Claude / Perplexity / Grok · plus any custom endpoints you add. A click never does nothing — every failure degrades to something visible (clipboard fallback, "couldn't extract" toast).

Options

content · header · position · theme (auto/light/dark) · bg · text · font (sans/serif/mono) · radius · items · endpoints · label · prompt (lead-in text sent to the LLM before the page's Markdown on the Open in… actions). theme: auto matches the host site, falling back to the OS, and stays live as the site/OS theme changes.

Add your own LLM target (enterprise/internal chat, self-hosted, anything that takes a prompt in the URL):

mount({
  endpoints: [{ label: "Open in Acme AI", href: "https://acme.ai/?q={q}" }],
});

{q} is replaced with the page's Markdown (or appended if the template has none); pages too long to inline fall back to a clipboard paste.

License

MIT