Where is my data stored, and how long is it kept?
Your inbox data is stored in Atlassian Forge SQL — Atlassian’s own hosted relational database infrastructure, within the Atlassian cloud. The only data sent outside Atlassian is sanitized comment text sent for AI classification (see AI & Privacy below).
How long items are kept
Section titled “How long items are kept”By default, items are kept for 90 days, after which they’re automatically removed. You can change this in Settings between 30 and 365 days — shorter retention means less stored data; longer retention means more history available to search and filter.
Items you resolve are also cleared automatically rather than kept indefinitely.
Managing storage manually
Section titled “Managing storage manually”If you’re approaching your storage limit, Krina automatically clears older resolved items first. You can also manage storage yourself from Settings → Data & Storage:
| Action | Effect |
|---|---|
| Clear oldest items | Removes items older than your retention setting |
| Clear read items | Removes everything you’ve already read |
| Clear all items | Empties your entire inbox — this cannot be undone |
Resolving items
Section titled “Resolving items”When you resolve an item, it disappears from your inbox immediately, and a confirmation appears with a 10-second “Undo” option. If you don’t undo within that window, the item is permanently deleted. If you do undo, the item is restored to your inbox as if it was never resolved.
AI & Privacy
Section titled “AI & Privacy”Comment text is sanitized before being sent to the AI provider for
classification. Sensitive values — including email addresses, phone
numbers, credit card numbers, social security numbers, API keys,
tokens, IP addresses, URLs, and hostnames — are detected and replaced
with placeholder tokens (for example, an email address becomes
[EMAIL]) before anything leaves Atlassian’s infrastructure. The
original values are never sent to the AI provider and are never written
to logs.
Note on @mentions: The Atlassian display names of mentioned users are intentionally not redacted before classification. They are essential context — the AI needs to know who is being asked, not just that someone is being asked, to classify the action and scope correctly. If you’re reviewing this product from a security or compliance perspective, this is a deliberate design decision and not an oversight.
See our Privacy Policy and Security page for full details.