Leaving NovelCrafter with your whole codex

Whatever is prompting the move, a subscription you no longer want, an AI workflow that stopped fitting how you write, or a bible that needs more than a codex, NovelCrafter deserves credit on the way out: it does the honorable thing and lets you leave. The export is real, it is yours without asking, and it contains nearly everything.

The word doing the work in that sentence is nearly. We took a filled-in NovelCrafter project, exported it in every flavor, and read every file, because an importer has to. This guide is what the export actually contains, the one feature that does not export at all, and how to move without retyping any of it.

What the export contains

A NovelCrafter export is a folder, and the layout is tidier than most:

• The manuscript as novel.md, novel.html or novel.docx, depending on the format you picked.
• codex.html, a rendered index of your codex. It is for reading, not for machines: names and text, stripped of the structure underneath. Ignore it; the real data is elsewhere.
• One folder per codex type (characters/, locations/, objects/, lore/ and so on), each holding a folder per entry. Inside each: entry.md with the entry's frontmatter and description, metadata.json with its id and relations, and notes.md if you used the Research tab.

Two structural facts worth knowing. Custom fields live only in entry.md's frontmatter, where text fields are plain values and dropdown or codex-reference fields are lists, with references recorded by name. And relations live only in metadata.json, as a list of entry ids held by whichever entry declared the relation, one-directionally, with no labels. Both survive a move; both would be easy to lose with a hand-rolled copy.

Export as Markdown

If you have the choice, export the manuscript as Markdown. It is the cleanest carrier of NovelCrafter's structure, and the structure is more mobile than it looks: the heading ladder shifts one level for every feature you used. Acts add a heading level, scene titles add another below chapters, and the plain scene separator arrives as * * * divider lines.

You do not need to remember any of that to import into PullAtlas, because the importer detects all four shapes. The html flavor imports too, and so does a novel.docx, so an old export you made months ago still works. Markdown simply leaves the least room for argument, in any tool you might ever point at it.

The thing that does not export

Progressions do not export. At all.

We added progressions to a test project deliberately, exported fresh, and diffed every file. Nothing progression-shaped appears anywhere: not in entry.md, not in metadata.json, not in codex.html, not in the manuscript. Whatever you wrote into progressions exists only inside NovelCrafter.

This matters more than it sounds, because progressions are the one place NovelCrafter records when something became true of a character, and that is usually the hardest-won information in a long project. If your arcs live in progressions, copy them out by hand before your subscription lapses, whatever tool you are moving to. Ours cannot import what the export does not contain, and neither can anything else.

How the codex maps

Drop the export folder, or a zip of it, on the PullAtlas import surface. Detection recognizes a NovelCrafter export outright, and nothing is created until you approve the preview.

• Types follow the frontmatter, never the folder name: characters become characters, locations become locations, objects become items, and lore, subplots and other land as notes until you say otherwise. Every type group in the preview has its own mapping dropdown, because only you know that your "lore" entries are actually factions.
• Aliases come across, and tags arrive as a Tags field.
• Custom fields become custom fields, with codex references carried by name, and the field names appear on the entry form from then on.
• "Do not track mentions" becomes "Don't highlight in the editor", the nearest honest equivalent.
• Research notes ride along under a Notes line in the entry's About text, with any external links.
• Relations become connections. NovelCrafter holds them one-directionally; the import joins them into proper two-way connections, deduplicated.

The manuscript comes too

The manuscript imports beside the codex, split into real chapters whatever combination of acts and scene titles you used. Scene separators become actual scenes, bold and italics become real formatting rather than asterisks in your prose, and the book's title comes from the export's own title line. Acts are flattened into the chapter list for now, and the import report says so rather than leaving you to notice.

If you would rather bring the bible into a book you already started here, the same import can merge instead: it matches entries by name and alias, fills only fields you left empty, never overwrites anything you wrote, and shows you every decision first.

After the move

The import ends with a report, not a shrug: entries by type, every rename, every relation it could not place, everything shortened to fit, and the progressions line above stated plainly, because a move that loses something silently is worse than one that tells you.

Then two things pick up where progressions left off. The free check scans your imported chapters for recurring names that have no bible entry yet and offers each one back to you. And the timeline, the part NovelCrafter could not hand over, rebuilds either by hand as you write, or from the AI pass that reads your manuscript against the imported bible and proposes chapter-anchored states, using your own AI key, only if you ask it to.