What it is: Atlas can now restore individual collections or databases directly into a live, running cluster through the UI or API from either scheduled/on-demand snapshots or point-in-time continuous cloud backups. Restores can overwrite existing data or create new databases and collections under custom names, with configurable handling options for indexes.

Who it's for: Teams needing targeted recovery instead of a full-cluster restore, such as fixing an isolated collection error, restoring one tenant's data in a multi-tenant deployment, pulling a subset of collections for analytics, or comparing an older version of a collection against the current one.

Why it matters: Previously, targeted recovery required restoring a full snapshot to a temporary cluster, extracting the required data with mongodump, and writing it back into Atlas with mongorestore. Atlas now manages this workflow natively, reducing manual effort and disruption to unaffected data and applications. The restore status page also provides progress and index-build details for each collection.

How to get started: From the Snapshots tab, select "Restore" or "Point In Time Restore," then choose the specific databases or collections you want to restore. See the documentation below for supported configurations and limitations.