The other two migrations on this site start with an export. This one doesn't.

Coming from Eagle means getting your files back out of a library that copied them. Coming from PureRef means accepting that the layout has nowhere to go because nothing reads a .pur scene. Coming from refern is a different shape of problem, because refern was never holding your files hostage in the first place — its own site states plainly that "normal workspaces read the files already in your folder without moving or duplicating them," which is the same architecture Coii Ref uses. Two index-based apps pointed at the same folder are not fighting over who owns the pictures. The only real question is what happens to everything you did inside refern once you stop paying for it, and that is what an importer is actually for.

What carries across, and what does not

The pattern here is close to the mirror image of the Eagle migration. There, almost nothing about metadata survives because Eagle's format was never read by anything on the Coii Ref side, but the files themselves need an export step first. Here, the files needed nothing, and the importer was actually built to read the one format that matters — because refern is the app Coii Ref's own import history started from.

Why there is no export step

It is worth confirming this rather than assuming it, since the whole shape of this migration depends on it. refern's own site describes its default behaviour today as reading files where they already sit — "without moving or duplicating them" — and separates that from its Eagle importer, which does copy on the way in and then uses linked copies for any additional folder placement. That second behaviour is specific to bringing files in from Eagle; it is not how a normal refern workspace treats a folder you pointed it at directly; So if you have been using refern the ordinary way, your images were never inside a container that needed prying open. They are exactly where Finder already shows them, and a Coii Ref workspace can be pointed at that same folder today.

The one place this gets more complicated is a refern workspace built partly from an Eagle import, where linked copies let one file appear in several folders at once without duplicating it. Coii Ref has no concept of a linked copy — every file it indexes has exactly one location — so a folder view that depended on that trick shows the file in its one real place and not in the others. That is worth checking before you start, not after: open the folders a linked copy used to appear in and confirm which one actually holds the file.

Before you start

Check whether any of your organisation depends on linked copies. This is the one refern-specific feature with no honest equivalent, so it is worth finding out up front how much of your workspace actually uses it rather than discovering a folder looks thinner than expected after the move.

Decide what to do about graph connections you actually used. If the relationship graph was decorative — connections you never went back to click through — losing it costs nothing. If you genuinely navigated your library by following links between images, that habit has no destination in Coii Ref, and it is worth knowing that before you commit rather than after.

Know which tags came from the auto-tagger versus your own hand. Both kinds come across identically, which is a feature for the migration and a small loss of information generally — you will not be able to tell, six months from now, which tags a human chose and which a model suggested and you accepted. If that distinction matters to you, it is worth exporting refern's own tag list as a record before you run the importer.

Note whether any workspace database or thumbnails live on a separate drive. refern's own site describes keeping a large library on a slow drive while its database, thumbnails and backups sit on a fast one. Coii Ref does not split a workspace that way — everything it writes lands beside the folder it indexes — so if your refern setup depends on that separation for performance, the new workspace will behave differently on the same hardware, and it is worth knowing that before rather than after the first big index finishes.

The steps

1. Install Coii Ref and point a new workspace at the same folder refern already indexes. Thirty days, every feature, no card and no account. Because neither app copies files, this step touches nothing that refern is using — both can read the same folder at the same time.

2. Let the first index finish. Coii Ref reads each file's dimensions, EXIF and colour and writes a thumbnail and a proxy beside it, entirely separate from whatever refern has already written for its own database elsewhere.

3. Run the importer. This is the step that actually moves anything: it reads the refern library and carries ratings, tags, notes and canvases across, matching everything by each file's relative path inside the folder. If you have reorganised folders inside refern since you last touched a given file, that file is the one the path match can miss — worth a quick before-and-after check on anything you moved recently.

4. Compare the two libraries while both still exist. Nothing about running the importer changes or deletes anything in refern, so this is safe to do slowly, checking a handful of folders by eye before trusting the rest.

5. Rebuild only what the table above says does not carry. For most libraries that is a short list: graph connections you actually used, and any folder view that depended on a linked copy. Everything else needs no manual work at all.

What you are giving up

Say this plainly, because refern remains the more capable app on almost every axis that is not price or platform. Its on-device auto-tagger suggests tags, reads text out of an image and describes what's in it, locally, which nothing on the Coii Ref side replaces going forward — the tags you already have carry across, but no new ones get suggested. Its search finds an image by describing it in words no tag captures, which Coii Ref's literal query line cannot do. The relationship graph, browser extensions for three browsers, linked copies, and imports from Eagle, PureRef and Allusion are all real capabilities with no equivalent here. And it runs on Windows and Linux as well as macOS, at $35 once against Coii Ref's $19 — a real difference in reach and a smaller one in price.

What you get instead is the same $19-cheaper, narrower app the rest of this site is honest about: canvases as plain JSON files rather than entries in a library, a workspace that is fully one folder rather than a database that can live on a separate drive from your images, and a search that answers only what is literally true of a file rather than what it looks like. If the auto-tagger or the meaning search is why you opened refern in the first place, this migration trades away the actual reason you were there. If your library is already well-tagged by hand and the graph was never load-bearing, the move costs an afternoon and not much else.

A worked example

Say your refern workspace covers about twelve thousand images across four project folders, tagged consistently for the last year, with a dozen moodboards built from the better-rated ones and a handful of graph connections between recurring subjects you never actually revisited.

Point a new Coii Ref workspace at the same parent folder and let the first index run — for twelve thousand files that is minutes, not hours, since nothing is being copied, only read. Run the importer once indexing finishes. Ratings, tags and notes land on every file the path match succeeds against, which for a folder structure that has not moved recently is effectively all of them. The dozen moodboards arrive as .coii-canvas files, already arranged, in whatever folder you kept them in.

Spend the next few minutes checking two things: whether any images used a linked copy to appear in more than one folder — if a handful do, decide whether the duplicate placement is worth a real file copy or was cosmetic — and whether the graph connections were actually navigational habit or just there. For most people the second answer is "just there," and the honest cost of this migration turns out to be smaller than expected: an afternoon, mostly spent confirming the importer did what it says rather than fixing anything it missed.

How to verify nothing was lost

Do this before letting a refern licence lapse, not after — nothing about running the importer changes refern's own library, so there is no rush.

Compare rating and tag counts on a sample of files, not the whole library. Open a handful of images in both apps and confirm the ratings and tags match; a path-matching importer either gets a file right or misses it visibly as untagged, so a spot check catches a systemic problem quickly.

Open every imported canvas at least once. Confirm the arrangement looks like the original moodboard rather than a placeholder — this is the one part of the migration doing the most work, and it is worth the two minutes per board to actually look.

Search specifically for what a linked copy used to supply. If you know a particular image used to show up under two different folders inside refern, check whether it still does in Coii Ref. If it does not, that is expected per the table above, not a bug to chase.

Keep refern installed until the graph question is settled. If it turns out you do use the relationship graph more than you thought during the comparison period, that is worth knowing before you commit to a narrower tool rather than after.

Worth reading next: the full Coii Ref vs refern comparison goes through every row refern wins, which is most of them, at more length than a migration page has room for. Best refern alternatives is the wider round-up if Coii Ref turns out not to be the right fit once you have read the concession above. Does refern work offline and what happens to a library if the app behind it stops shipping are both worth reading regardless of which way this goes. If Eagle or PureRef are also in the mix, migrating from Eagle and migrating from PureRef are both a noticeably harder version of this same move. Otherwise, the 30-day trial is the fastest way to find out whether the importer above actually carries across what your own library needs it to.