Key highlights
- Abandoned carts: a cart left untouched raises an Abandoned event that a job can act on, such as a recovery email.
- A faster file manager: large folder trees open one level at a time.
What’s new in Core dna Platform 2026.3.1
Abandoned Cart Detection and Faster File Manager
Carts left untouched are flagged as abandoned, so a job can send recovery emails. The file manager opens large folder trees much faster.
Changes in behaviour
- Abandoned cart detection is on for every site with carts, with a threshold of one hour. Review the threshold in the Carts settings before building jobs on the event.
Abandoned cart detection
Carts left untouched are now detected automatically. When a cart has a customer email and items, has not reached checkout, and has not changed for longer than the Abandoned Threshold in the Carts settings (in seconds; an hour by default), it raises an Abandoned event. A job on that event can send a recovery email or anything else you need. Carts of guests and of signed-in customers both count.
A faster file manager
The file manager now opens large folder structures much faster: it loads the top-level folders first and fetches each folder's contents when you open it, instead of loading the whole tree up front.
Other improvements and bug fixes
- Blog, post and comment counts shown in the admin are now calculated live from the actual data, so they are always accurate instead of relying on stored totals that could fall out of sync, and they are computed more efficiently when listing multiple blogs or posts.
- Scheduled tasks now respect their pause and recurring-until settings. Paused tasks no longer run until their pause period ends, and tasks past their recurring end date stop executing as expected.
- Fixed an issue where saving an FAQ page without selecting any questions caused a server error. The form now displays a clear validation message prompting you to add at least one question.
- Resolved an issue where AI agents and command-line tools that use a local loopback address (such as 127.0.0.1 or localhost) for their redirect could not complete the connector authorization flow. These connections now work reliably.
- Fixed an issue where connecting an external AI agent by URL could be incorrectly rejected when its host resolved to a different address on repeated DNS lookups, for example behind a load balancer or round-robin DNS. Valid connections are now accepted while security checks remain in place.
- Fixed an issue where the user profile page in the admin could fail to load with a server error when it contained settings that use the select or multi-select-tree editors. The profile page now opens correctly.
Go deeper
Was this page helpful?
Your answer goes to the docs team.