Markdown Editor

The commonly-written core plus GitHub-style tables: ATX headings, bold and italic, inline code and fenced code blocks with a language tag, links and images with optional titles, ordered and unordered lists nested by indentation, blockquotes that can hold any other block, horizontal rules, and the two-trailing-spaces line break. Setext headings and reference-style links are deliberately absent — they are rare in the wild.

By design. This renderer escapes everything before reading any markdown, so a pasted <script> renders as the characters, never as an element — and a [link](javascript:…) loses its destination while keeping its text. That safety is what makes the live preview possible at all: the output can be injected into the page because it cannot carry code.

Underscore emphasis only applies at word boundaries, matching how markdown is actually written: identifiers full of underscores survive untouched, while _a phrase set off by spaces_ italicises. Asterisks work anywhere, so *this* is the operator to reach for mid-word.

Copy HTML puts the rendered fragment on the clipboard — the h1/p/table elements only, ready to paste into a CMS, an email editor or a template. Download .html wraps the same fragment in a minimal standalone page with readable styling, so the file opens correctly on its own.

End the line with two or more spaces. A single newline soft-wraps into the same paragraph — the behaviour that surprises everyone once — and a blank line starts a new paragraph. The two-space rule is markdown's own, invisible in the source, which is why editors that trim trailing whitespace quietly break it.

No. Rendering runs entirely in this page — drafts, notes and documentation are exactly the kind of text that should not travel, and here it never does. The page keeps working with the network disconnected.