keelapps/Curator
For the admin holding a list of three hundred pages
Change three hundred pages at once — and put them back if it was wrong
Curator filters a Confluence space or a page tree, then re-labels, moves or archives what it found. Every batch is previewed page by page before a single write, and every batch but archiving can be undone as a whole — with each page guarded, so a colleague's edit is never overwritten by an undo. Nothing leaves Atlassian.
Free for sites up to 10 users. Above that it is priced per user per month and billed by Atlassian — see pricing on the Marketplace.
Confluence Cloud has no bulk page or label management. Retiring a label across three hundred pages, or moving a project tree into another space, is manual work — and the requests for it have sat at Gathering Interest for years: CONFCLOUD-7681 (rename, merge and delete labels), CONFCLOUD-17373 (mass add/remove labels), CONFCLOUD-73945 (bulk move pages between spaces, where the official answer is “use the API”), CONFCLOUD-78606 and CONFCLOUD-70744 (bulk archive).
The Marketplace category is fragmented and its leaders are small: page managers that do not touch labels, label managers that do not touch pages, and — the part customers complain about — nothing that can be undone.
What it does
Pick pages precisely
Per space or per page tree: title contains, updated before or after, author, has all of these labels, has none of these, include archived. The result is counted before anything happens.
Change them in bulk
Add, remove or replace labels. Move pages under a different parent page, or into another space. Archive them. One filter, one action, one batch.
A dry run, every time
Before a single write: how many pages change, the first fifty with the exact per-page edit, how many are skipped and why — every skip reason gets its own sentence, never a shrug — and whether the operation can be undone.
Undo the batch, not the page
Every run records a per-page inverse, and undo replays it. Each page is guarded: if somebody else changed it in the meantime, Curator leaves it alone and reports it rather than overwriting their work.
Govern labels site-wide
Browse every shared label with a usage count, then rename, merge or delete it across the whole site. Site admins only. Personal favourites are never listed and never touched.
History you can hand over
Who ran what, when, over how many pages, and whether it can still be taken back — exportable as CSV. An undo is itself an operation in that history.
Retiring a label, end to end
-
Find out what carries it
The label catalogue lists every shared label on the site with a usage count and the time that count was taken.
deprecated-2024turns out to be on 312 pages across nine spaces. -
Read the dry run
Rename it, and before anything is written Curator prices the batch in API points, names how many pages would change, lists the first fifty with their exact edit, and reports every page it would skip with its own reason.
-
Run it, paced
Confluence's hourly API quota belongs to the app and is shared across every site running it, so the batch paces itself: when this site has used its share it pauses and resumes in the next window rather than draining the pool for somebody else's tenant.
-
Take it back if it was wrong
One click replays the inverse on every page. Any page a colleague edited in the meantime is skipped and reported, not overwritten — and the undo appears in the history as its own operation.
Against what you have today
| Question | Native Confluence | Page or label managers | Curator |
|---|---|---|---|
| Re-label three hundred pages at once? | No bulk editing | One or the other, rarely both | Pages and labels, one batch |
| Rename or merge a label? | No rename endpoint at all | Sometimes | Site-wide, previewed |
| See exactly what a batch will change first? | — | Counts, if anything | Page by page, with every skip reason |
| Undo the whole batch afterwards? | — | Usually not | Per-page inverse, guarded |
| Move pages into another space? | Manual, page by page | Some | In the same batch |
| Does anything leave Atlassian? | — | Depends on the vendor | Never |
Limits we state up front
Two of these come from Confluence itself. Curator says so in the interface — in the preview, on the confirm button and in the history — rather than letting you discover them halfway through a batch.
Archiving cannot be undone. Confluence has no restore API:
unarchive appears nowhere in either REST specification, and
CONFCLOUD-75065 is still Gathering Interest. Archive is marked
non-reversible everywhere it appears. Everything else Curator does is
reversible.
Renaming a label is not an atomic operation. Confluence has no rename or merge endpoint — a label is a string on a page, not an entity. Curator does the only thing that exists: remove the old label and add the new one on every page carrying it. Which is precisely why the dry run, the resumable checkpoint and the batch undo matter most here.
Label usage counts are derived, not reported. Confluence's label endpoint returns names and ids and nothing else, so Curator counts by walking each label's pages in the background and stamps every row with when it was counted.
Attachments are out of scope, deliberately. For those, see Satchel for Confluence.
Availability
Curator is live on the Atlassian Marketplace. It is built and deployed on Atlassian Forge and installs into your Confluence Cloud site straight from the listing.
Licensing is handled by Atlassian. Sites with 10 users or fewer pay nothing at all, with every feature included; paid tiers are priced per user and taper as the site grows.
Questions, bug reports and feature requests: raise a support request.
Found a security issue? Please report it privately rather than in a public channel.