You'll noticeTier: how much it should matter to you
4Useful: my rating, 1 to 5
3Signal: worth watching, 1 to 5
Cloud SyncArea: what it touches
ImprovementsKind: in v2.1.281,
What probably matters to youSection of the release
What
A cloud session runs Claude Code on a remote machine, inside a container (an isolated computing environment), and directory sync keeps its copy of your files in step with your own machine. When that container was replaced, the checkout (the copy of your project Claude works in) used to start over from the session's starting state and wait for your machine to send files again. Claude's earlier work in the container is now restored where possible, and Claude is told how much came back.
There are now these outcomes:
Full restore: Claude's commits, staged changes and working files are back as they were at the end of its last turn.
last_turn: work is restored only up to the end of the last completed turn.
partial: work is restored only up to an earlier turn, and a reason is given.
Failed restore: Claude is given the reason and what happens next, which is resends, refused, taken, or that your changes have not yet been taken in.
The report shown to you gets matching lines. A new telemetry event, dir_sync_git_checkout_restored, is logged with an outcome field, and the existing dir_sync_git_checkout_recreated event now carries a why field.
Why
A replaced container no longer has to mean losing what Claude had done. Claude is told plainly how far back its work was restored, so it knows which of its changes are still there and which it has to redo.
How sure we are
One source agreesOne thing we can check says the same as this entry.
Anthropic's release notes agreeImproved screen-reader output in /mcp: a disabled server is read as "off" instead of "pending"
