Repository maintenance
Over time, a sync repository can accumulate history, references no longer in use, or anomalous states of the local clone. The maintenance operations tidy things up without ever putting your data at risk. You find them in the System tab of the dashboard (and in the corresponding commands), each with its own repository card — in the Team edition there are Core and Memories.
The local clone is just a cache
First of all, a guarantee that makes all these operations safe: the local clone of the repository is
a derived cache. The source of truth is the ~/.claude folder, where your Claude Code
configuration lives. That’s why clearing and re-cloning the repository loses nothing: it simply
rebuilds the cache from scratch.
The operations
Every operation always asks for in-app confirmation before running.
Reset / Reinitialize
Clears the local clone and re-clones it from scratch. It’s the safest operation and often resolves many anomalous states: because the clone is a derived cache, rebuilding it involves no data loss.
Compact history
Reduces the repository’s size by compressing the history into a single commit. The previous history doesn’t disappear: it remains reachable through a recovery ref kept for the purpose.
Clean
Removes the “unmanaged zones” from the repository: content that’s no longer synced and is therefore no longer needed in the repository.
Change repo
Migrates the history to a new URL. Useful when the repository is moved or renamed, or when you want to adopt a different git host.
When to use them
| Situation | Operation |
|---|---|
| The sync status stays anomalous and doesn’t resolve on its own | Reset / Reinitialize |
| The repository has grown a lot and you want to lighten it | Compact history |
| Content that’s no longer synced remains and you want to remove it from the repository | Clean |
| You’ve moved or renamed the repository, or you’re changing git host | Change repo |
See also
- Privacy and security — why idempotency and the absence of force-push protect your data.
- Team edition — maintenance on two repositories, CORE and MEMORIES.
- Troubleshooting — diagnostics and anomalous states.