Bringing a backlog in
Project → Settings → Import from a file. Pick a CSV, check what each column means, read the summary, import.
Nothing is written until the last step. The preview is a real run against the server with the writing switched off — the same code — so what it promises and what lands cannot drift apart.
What it reads
Section titled “What it reads”Exports from Jira, Linear, Plane, OpenProject, Trello and Todoist are recognised by shape. Any other CSV works too; the mapping is guessed from the header names, in English and German and in the words those tools actually use:
| Field | Headers it recognises |
|---|---|
| Title | Title · Summary · Subject · Name · Titel · Aufgabe |
| Description | Description · Details · Notes · Beschreibung |
| State | Status · State · Stage · Column · Zustand |
| Type | Type · Issue type · Work item type · Art · Vorgangsart |
| Priority | Priority · Severity · Priorität |
| Assignee | Assignee · Assigned to · Owner · Bearbeiter |
| Labels | Labels · Tags · Components · Schlagworte |
| Due date | Due · Due date · Deadline · End date · Fällig |
| Estimate | Estimate · Story points · Schätzung · Aufwand |
| Original ID | Key · Issue key · ID · Ticket · Nummer |
| Parent | Parent · Epic · Epic link · Übergeordnet |
| Blocks / Blocked by | Blocks · Blocked by · Depends on · Predecessor |
Every guess is a dropdown you can change, and anything unmapped is ignored.
It copes with what real exports actually contain: semicolon separators from a German Excel, tabs and pipes, quoted fields containing commas and newlines, a byte-order mark Excel writes and never mentions, and short rows with the trailing empty columns left off.
How values land
Section titled “How values land”| Priority | Highest, Blocker, Major, Trivial and the German words all map onto Kolibri’s five. Anything it cannot read is reported and the priority left alone |
| State | Matched by name against the project’s workflow, case-insensitively. No match → the project’s default state, and a line in the report |
| Type | An issue type from another tracker arrives as a label, matched against labels the project already has, so Bug lands on an existing bug rather than beside it |
| Assignee | By email, full name or first name. No match → unassigned, and a line in the report |
| Labels | Split on , ; or ` |
| Dates | 2026-12-31 and 31.12.2026. 01/02/2026 is refused — that is two different days in two different countries, and guessing wrong moves a deadline five weeks without saying so |
| Original ID | Appended to the description as Imported from PROJ-417, because that is what somebody searches for later |
Parents and blocking links are resolved on a second pass, once every row exists — so a file where a child appears before its parent works.
A row it cannot fully read
Section titled “A row it cannot fully read”It imports it, and says what it could not read.
A row with a title and four unreadable columns is still a task somebody wanted; refusing the file over it means the file gets fixed in a spreadsheet, which is exactly the work the import was meant to avoid. Every problem names the spreadsheet row number — the header is row 1, as Excel counts — and what it could not make sense of.
A row with no title is skipped, and that is also reported. A task without a title is nothing.
Limits
Section titled “Limits”- 5 000 rows per file. More is refused rather than run for ten minutes behind a spinner.
- No comments. Parents and blockers come across; comments need the JSON format below.
- One project at a time. The target is the project you started from.
- No undo button — but an import is ordinary tasks. Filter to them, select them all, delete them together. And if you delete one too many, Settings → Data has it.
- CSV only from other tools. All of them export it.
Moving between two Kolibri instances
Section titled “Moving between two Kolibri instances”Project → Settings → Export as JSON writes the project as a document, and Import a JSON export reads it back. That round trip carries what CSV cannot: comments, relations, custom fields, cycles and pages.
Two things it is not. It is not a backup — kolibri backup copies the database, which has to
be exact — and it is not the way to duplicate a project inside one instance, which is
Copy this project. See across projects.
If somebody in the export does not exist on the receiving instance, their work arrives unassigned and the report names them.