What you get
Every server with Server Backup gets at least 7 daily and 7 weekly restore points. Add more dailies (up to 30), more weeklies (up to 12) or monthly points (up to 12) for ₹9 each per month. A weekly or monthly point is the daily backup of that day kept longer, so it never doubles your storage.
Server Backup and Vault Protect are available to accounts that have made at least one payment; add funds to your wallet first if the account has only used promo or welcome credit. Backups run at the hour you choose (IST) and are billed hourly from your wallet like the server. Once a day you can also take a manual backup, kept 24 hours on top of the schedule.
Pick it when you order (the Build page or the order page): on an account that has not paid yet it starts automatically after the first payment, held up to 30 days.
Storage: 100 GB free per account
Restore points are incremental: weekly and monthly points are full copies, daily points store only the blocks that changed since the last one, all compressed and encrypted with a key for that server. Restore points are stored in an S3 backup bucket in your account. The first 100 GB is free and shared by all your servers; above that you pay the S3 rate in 100 GB steps. When backups need more room, the bucket grows by one step and we email you why; when old points expire it shrinks again.
Restoring
Restore in place rolls the server back to a point: a copy of the current disk is kept 24 hours first, so you can undo it. Restore as a new server builds a separate server from a point with a new IP, handy for testing. Or download the disk image. All restores are free.
Browse files opens a restore point read-only in the panel: walk the folders or search by name, tick files and download them as a zip (link valid 24 hours). Your server is not touched.
App-consistent backups
With the QEMU guest agent running inside the server (`apt install qemu-guest-agent`, `dnf install qemu-guest-agent`, or the virtio-win guest tools on Windows, which use VSS), each backup pauses disk writes for a moment so databases and open files are captured cleanly. Without it a point is crash-consistent: the same state as after a sudden power cut, which journaling filesystems and databases recover from on boot. The panel shows which one each point is.
Restore tests
Every month we boot a sample of backups in an isolated network to prove they restore; servers that pass show Verified restorable. You can also test one yourself once a month, free: we boot the latest point next to your server, never touching it, and give you a console for an hour.
Through the API and MCP
`protection: backup` on `POST /api/v1/services` (MCP `order_servers`) orders it with the server. `PUT /api/v1/services/<id>/backup` turns backups on or changes the schedule, `GET /api/v1/services/<id>/backup/points` lists restore points and `POST /api/v1/services/<id>/backup/points/<point>/restore` restores one. The MCP tools are `set_server_backup`, `list_restore_points`, `backup_server_now` and `restore_backup_point`.