hbkit: open Synology Hyper Backup (.hbk) files without a NAS

Before you restore

Find out first if it will work

hbk doctor reads the archive's layout, then checks a sample: it rebuilds files from the selected backup version and checks every chunk in the sample. A passing sample does not check every file.

A wrong password is rejected before any data is read.
A layout hbkit doesn't know is refused, never guessed.
On an encrypted archive, it shows how many backup versions' keys it found and used.

hbk doctor

$ hbk demo/synth.hbk doctor
          chunk codec : lz4
            encrypted : no
               shares : Documents, Photos
      backup versions : 3
      reading version : 3 (2026-10-28 08:33:20, Complete) - newest complete

  PASS  virtual_file layout known  (56 B)
  PASS  chunk_index layout known  (29 B)
  PASS  bucket layout known  (32 B)
  PASS  rebuilt 6 sampled files, all chunks verified  (6 ok, 0 failed)

  VERDICT: recoverable. Sampled files rebuilt byte-exact and checksum-verified.

Real hbkit output, shortened. The archive is a small one that hbkit's own tests build, not one from a NAS.

Made for the day the NAS is gone

When the NAS is gone, the backup is all you have left. Synology's Hyper Backup Explorer can open it, but its Mac and Linux versions are built for Intel (x86) only. Apple says Rosetta, which runs Intel apps on Apple silicon, stays available for all apps only through macOS 27. hbkit runs natively on Apple silicon and on Linux servers with no screen. hbkit is an open reader for your own backups. FORMAT.md is a public description of the .hbk format, written so archives stay readable.

Safe by design

It won't hand you a broken file

Checked on the way out

Every chunk is checked against the MD5 and CRC32 stored in the archive. The worst case is a loud failure that names the file, never a file that is quietly wrong.
Read-only

hbkit never writes to the archive. If it is your only copy, copy the whole .hbk folder to another disk first and run hbkit on the copy.
Resumable and crash-safe

Each file is written to a .part file and renamed when it is complete, so a stopped run never leaves a cut-off file behind. Run the same command again and finished files are skipped.
The original layout

Files come back under their share and original path, with their original modification times. Synology's thumbnail and metadata folders stay out of your restore.

Browse

Tick folders, then restore

hbk-tui shows the whole backup as a tree, with each folder's size and file count. Tick what you want, choose where it goes, and press r.

It warns you before you start if the selection won't fit.
While it runs: live throughput, time left, the files it just finished, and a log of any failures.
The file list is cached per archive. After the first open, a backup of 1.1 million files opens instantly.

The side panel: 52,386 files and 244.4 GB selected in 2 items, the destination with 1.6 TB free, a Start recovery button, and the archive's totals — The selection panel warns before you start if the files will not fit.

Cloud copies

Restore straight from a bucket

Mount the bucket with rclone and point hbkit at it. hbk mount sets the rclone options that matter and loads the folder listings up front, so the wait happens once, at the start. Your cloud keys stay in rclone's own config.

Browsing and search use a local index, with no network at all.
Checksums cost nothing here: decompressing and checking took 0.7% of the time.
A copy on a local disk is still much faster. Try one small folder first.

A 3 TB archive in Cloudflare R2

Before tuning, then with hbk mount.

Mount and load the folder listings about 12 minutes before 16 s
Open the archive 9 min 21 s before 1.8 s
Restore a 10 KB file 584 s before 11 s

Command line

Or script the whole thing

Every command takes a pattern over the archive's own paths, which start with the share name. Output keeps the folder tree and the original dates.

-j sets how many worker processes run at once. Use 16 on a network mount.
--strict re-checks the recorded checksums when resuming, so a file of the right size with the wrong content is caught.
Plain output and a non-zero exit code on any failure, for cron jobs and scripts.

zsh

# restore a folder, keeping the tree and dates
$ hbk /Volumes/Backup get "/photo/2023/*" ~/restore

# check every file without restoring it
$ hbk /Volumes/Backup verify "*" -j 16

# find a file by name
$ hbk /Volumes/Backup list invoice

# info prompts for the password, or use HBK_PASSWORD
$ hbk /Volumes/Backup info

What it works with

hbkit has been checked byte for byte against a handful of real archives, and doctor exists so you can check yours in seconds, not halfway through a restore.

Where it lives

Any folder your computer can open: an internal or USB disk, or a mounted NAS share
S3-compatible buckets such as Amazon S3, Cloudflare R2 and Backblaze B2, through an rclone mount

Computers

macOS, Apple silicon and Intel
Linux
Python 3.10 or newer, and liblz4

On Windows, CI restores test archives byte-exact on every change. A backup from a real NAS has not been restored on Windows yet.

Ways to install

Homebrew, on macOS and Linux
pip or pipx, from PyPI
An Arch PKGBUILD and a Nix package, in the repository

Install

One command to install, one to check the archive, one to restore.

Install hbkit
```
$ brew install YordiLorenzo/tap/hbkit
```
Homebrew installs liblz4 with it. With pip or pipx, run pip install hbkit and install liblz4 yourself: brew install lz4, apt install liblz4-1 or pacman -S lz4.
Check the archive
```
$ hbk /Volumes/Backup doctor
```
Give it the .hbk folder, or any drive or folder that holds one. For an encrypted archive, add -p and the password, or set HBK_PASSWORD.
Restore
```
$ hbk-tui /Volumes/Backup
```
Or from the command line:
```
$ hbk /Volumes/Backup get "/photo/*" ~/restore
```
If it stops, run the same command again to carry on.

Every command, network mounts and performance tuning are in the README.

Questions

The README goes deeper, and you can always open an issue.

What is a .hbk file?

It is the folder that Synology Hyper Backup writes to the backup destination, for example nas_1.hbk. Inside are folders such as Config and Pool. Your files are not kept there as plain files. They are split into small compressed chunks, so a file browser cannot open them. hbkit rebuilds the files and checks every chunk.

How is hbkit different from Hyper Backup Explorer?

Hyper Backup Explorer is Synology's own desktop app. Like hbkit, it lets you pick a backup version. It can also open backups in Synology C2. C2 is not tested yet with hbkit. hbkit runs natively on Apple silicon Macs and on Linux servers with no screen. It has a command line with documented exit codes, a terminal browser and the doctor check. It checks every chunk against the MD5 and CRC32 stored in the archive. Use Explorer for C2, or for a graphical app on Windows. Use it also for key-file access and older RSA-encrypted archives.

Do I need a Synology NAS or account?

No. hbkit reads the archive's files directly. It doesn't need a NAS, Synology software or a Synology account. After installation, it works offline for a local disk.

Can it damage my backup?

No. hbkit opens the archive read-only and never writes to it. Restored files go only to the destination you choose.

My backup is encrypted. Will it work?

Yes, with the password. hbkit checks the password against the public key stored in the archive, so a wrong one is rejected straight away. Older archives encrypted with RSA, and access with the exported key file, aren't supported yet.

Can I restore straight from S3, R2 or B2?

Yes, through an rclone mount, and hbk mount sets it up for you. A restore reads many small pieces, each one a trip over the network, so it is much slower than a local disk. If you can, copy the archive down first.

Does it work on Windows?

On Windows, CI restores test archives byte-exact on every change. A backup from a real NAS has not been restored on Windows yet. Install it with pip, then set HBK_LZ4 to the full path of an lz4.dll (CI builds one with vcpkg install lz4:x64-windows). For a restore that matters, use a Mac or a Linux machine, or Hyper Backup Explorer on Windows.

Can it read a backup in Synology C2?

Not tested yet. hbkit has not been tried on a backup kept in Synology C2 Storage. Synology's Hyper Backup Explorer can browse and download C2 backups, so use it for C2 for now.

How fast is it?

Usually as fast as the disk allows. From a USB hard disk that reads 92 MB/s at best, a restore ran at 58 MB/s. From Cloudflare R2 it reached about 9.5 MB/s with 16 workers.

Can it restore an older backup version?

Yes. hbkit reads the newest complete version by default; hbk <archive> versions lists them all, and --version N or --version @2026-10-15 picks another for list, get, verify, doctor and the TUI.

What if doctor fails on my archive?

Please open an issue and paste the output. Fixes in 0.4.5 and 0.4.6 came from people whose archive wouldn't fully restore: one had gaps in its index files, the other was an encrypted archive with seven backup versions.

Is hbkit made by Synology?

No. It is an independent open source project. The format was worked out from real archives, and is written up in FORMAT.md for anyone who wants to build on it.

Get your files back

Free and open source, for macOS and Linux. Run doctor first, then restore what you need.

$brew install YordiLorenzo/tap/hbkit

Release notes

Version 0.5.1, released 2026-11-09.

An archive hbkit can't read

Every Synology setup is a little different, and each new archive teaches hbkit something. If doctor doesn't pass on yours, the output is enough to start from.

Open an issue How the format works

Your backup, without the NAS.

Find out first if it will work

Made for the day the NAS is gone

It won't hand you a broken file

Checked on the way out

Read-only

Resumable and crash-safe

The original layout