Prettier sets JavaScript, TypeScript, HTML, CSS, SCSS, Less, JSON, YAML and Markdown; SQL has its own formatter. Other languages keep their shape: comments and stray whitespace go, your indentation stays. In Python, docstrings count as comments (a body left empty gets a pass); triple-quoted strings used as values are kept. Everything uses a 4-space indent.
The inline long-lines switch keeps each statement on one line instead of wrapping at 80 columns, wherever a formatter runs. For languages without one, it joins bracketed argument lists and literals onto a single line, turning any comments inside into /* … */ form so they ride along; blocks and multi-line strings are left untouched.
Collapse comments folds your pinned notes down to a brass marker in the gutter: hover the marked line to read the note, click the marker to open it, and Fold to close it again. Code comments are unaffected.
Reading aids: the enclosing function’s signature stays pinned at the top while you scroll; Outline folds the proof to its top-level signatures; Wrap wraps long lines; spotlight dims everything except the function under the cursor; matching brackets share a colour, and the active pair gets a guide.
Explain asks Claude to describe the function under the cursor (or your selection, or the whole file if the cursor is outside any function) and pins the answer as a note you can edit, fold, or copy out. Inside claude.ai it needs no setup; as a standalone file it asks once for an Anthropic API key, kept in memory only.
To pin a note, click the + in the gutter for one line, drag the + across several lines, or select lines and press Add note. Notes live outside the code, so the stripper never touches them; they last until the page closes — Copy with notes takes them along as comments.
Clear the editor?
The draft will be discarded.
Explain needs an API key here
Inside claude.ai this works without one. Running the file on your own, paste an Anthropic API key. It stays in memory until the page closes and is sent only to api.anthropic.com.