Minimal escapes only the five characters that can break HTML: & → &, < → <, > → >, " → ", ' → '. That is enough for any UTF-8 page, in text and in quoted attributes.
Numeric also writes every non-ASCII character as its code point (é → é or é), so the result is pure ASCII. Named uses names like é € — for the HTML 4 set (Latin-1, Greek, arrows, math, typography) and numbers for the rest.
Decode understands all 2,000+ HTML5 names plus &#…; and &#x…;, the same way a browser does. Unknown names and broken numbers are listed above the output.
Shortcut: Ctrl/⌘+Enter copies the output. Link straight to a result with #decode=<text> or #encode=<text>.