Checkpoints & reset
Checkpoints and reset exist so that you and your agent can try things that might break the setup. They work per project.
Reset deletes every resource of the project: servers, volumes, networks, load balancers, firewalls, IPs, DNS zones, and anything that is protected against deletion. Checkpoints stay. The project is empty when the action finishes.
- In the app: the project’s Reset button, or ↺ next to a session in the menu bar. The app asks first.
- For an agent: the MCP tool
reset_project.
Checkpoints
Section titled “Checkpoints”A checkpoint saves the project’s resources with their names, settings and relations, plus a snapshot of every server’s root disk. Restoring it empties the project and creates everything again, servers from their snapshots.
| Action | In the app | MCP tool |
|---|---|---|
| save | Checkpoints › New Checkpoint, or ◆ in the menu bar | create_checkpoint |
| list | Checkpoints | list_checkpoints |
| restore | Restore on a checkpoint; the app shows what the restore deletes and creates | restore_checkpoint |
| delete | Delete on a checkpoint | delete_checkpoint |
A project keeps at most 20 checkpoints. Their snapshots count toward the project’s image quota.
What a restore keeps and what it does not
Section titled “What a restore keeps and what it does not”- Names, labels and private addresses stay the same. Resource IDs change. Tools that address resources by name, as Terraform does after a refresh, are not affected.
- Public addresses may change. DNS A and AAAA records in the project that pointed at the old addresses are rewritten to the new ones.
- Volume contents are not saved. Volumes come back empty, formatted as before. Keep data you need on the server’s root disk or in S3, or restore it from a backup.
- Snapshots are taken while servers run, so a server comes back as if its power had been cut at that moment. Databases recover the way they do after a power cut.
- Reverse DNS entries are not restored.
Checkpoints and Kubernetes clusters
Section titled “Checkpoints and Kubernetes clusters”A checkpoint covers the servers of a project, including the VMs a lab cluster runs on. For the state inside a cluster, use the cluster’s own backups: the Backups tab of a cluster shows its etcd snapshots and Snapshot Now takes one. See Kubernetes clusters.