Someone synced a bad folder over 400 files at 14:02. The files are still there. They are just all wrong.

The gap Microsoft leaves open

The missing middle is: select a precise subset by search, then roll only that subset back to a point in time. That's the product.

What it does

• Find the damage. Filter a library by modified-between, modified-by (people picker), path, extension, filename pattern (ransomware suffixes), size change, or version count. Filters combine; the match count updates as you type.
• Roll back per file, not per version number. Each file goes to its own newest version created before your chosen time — version numbers differ per file, so "version 12" means nothing across a set.
• Dry-run with five buckets. Restorable · created after the incident (flagged, not touched) · no prior version · checked out by someone else · already current. Counts by reason, never one number.
• Batches, throttling, resume. Built for 20,000-item libraries, honours Retry-After, and picks up where it stopped if you close the tab.
• Success and failure CSV, per item, before you leave the page.

What it will never do

Restoring a version creates a new current version and removes nothing; the old state stays in history. The product has exactly one write endpoint — restore file version by label — and no delete, purge, or version-removal call anywhere in the code. Same published-endpoint-list, same build-time check, same verifier as Undelete365.

Scope of the first release

Pre-order pricing

One-time, per tenant, no subscription — the same model as Undelete365. Pre-order buyers pay less and are first to get the key at launch.