BackDown
Early access · Linux first

Backup that thinks.

BackDown copies your drives like any good backup tool, then tells you what's on them: what's duplicated, what each folder actually is, where it came from, and which drives are safe to wipe. Every claim is backed by a hash.

You're on the list. We'll email you once, when it's ready.

That email address didn't work. Check it and try again.

One email when it's ready. No tracking, and we never share your address.

Six drives, one datasetFindings
Safe to reclaim
5.9 TB
duplicates, SHA-256 verified
Indexed
27.1M
files across 15.2 TB
  • Dovecot maildirmail.example.com · 2011–2020 · rsynced 31 May 202038 GB
  • Time Machine setMacBook Pro · 212 snapshots1.4 TB
  • iPhone backuplast written March 202364 GB
  • Git repositories143 repositories12 GB
  • Caches skippednode_modules, .cache, thumbnails310 GB
Seagate 2 TB✓ Safe to wipe

Every byte has a verified copy.

WD Passport✗ Copy first

212 GB exists nowhere else.

Example report. Headline numbers are from the real run that started BackDown; the entities are illustrative.
Where it came from

Built on 27 million files and a drawer of old drives.

BackDown began as a one-off job: consolidate years of old disks onto one archive and wipe the rest. The hard part was never copying bytes.

Five of six drives held generations of the same data. The most valuable thing on them, a mail server's archive, only turned up by reading the shell script that once backed it up. And 5.9 TB of the biggest drive was quietly duplicating itself.

No backup tool would have told us any of that. So we built one that does.

27M
files indexed, with every job resumable when a USB drive dropped off the bus
15 TB
of storage read and hashed, with source drives mounted read-only throughout
5.9 TB
of the biggest drive found duplicating itself
3
drives wiped, each only after every byte was hash-verified elsewhere
How it works

Plug in everything. Get answers, not a vault.

Backup tools answer "can I restore this file?" BackDown also answers "what do I have, what's duplicated, and what's safe to delete?"

Point it at every drive you own

Old laptops, NAS shares, the USB disk in the drawer. Sources are mounted read-only and tracked by serial number, never by /dev/sdX, so the wrong disk is never touched.

It indexes, hashes and recognises

Tiered hashing checks size first, then a quick sample, then a full SHA-256, so it only fully reads files that could be duplicates. If a drive disconnects or the machine reboots, you lose the file in flight, not the night.

You get a verdict for every drive

What's on it, what each folder is, what's duplicated across every source, and a verdict: safe to wipe, or copy these 212 GB first. Nothing is deleted without a verified copy.

Features

The intelligence other backup tools leave out.

Borg, restic and Time Machine are excellent vaults. BackDown adds the layer that tells you what you're actually storing.

Works todayIn developmentPlannedBackDown is pre-release, so each feature is marked with its status.
Works today

Dedup across everything you own

Not just inside one repository. Each new drive is checked against everything already indexed. Hardlinks and symlinks are accounted for, so reclaim figures are real.

Works today

Proof a drive is safe to wipe

Either every file has a SHA-256-identical copy elsewhere, or you get the list of files that don't. Every copy and deletion is journaled against a verified hash.

Works today

Survives bad hardware

Every long job resumes where it stopped. Flaky USB enclosures are remounted automatically, and jobs run one per spindle so a single disk never thrashes.

In development

Knows what things are

Maildirs, Time Machine sets, phone backups, home directories, disk images, git repos, SQL dumps and cloud exports, each with a date range and size. Regenerable caches are flagged and skipped.

In development

Knows where it came from

BackDown reads the rsync scripts, sync markers and backup metadata that record your data's history. It can say "rsynced from your mail server on 31 May 2020," not just "some files."

Works today

A web UI you'll actually open

Live progress, interactive size trees, duplicate tables and per-drive verdicts, served locally by the daemon, in light and dark.

Planned

Browse without restoring

Mailboxes appear as threaded mail, photos as a gallery, SQL dumps as schemas. Disk images mount read-only. Search every backup and drive at once.

Planned

Scheduled, encrypted backups

Content-defined chunking, zstd compression and client-side encryption, verified on a schedule, to local disks, SFTP, S3 or B2.

Planned

A chain of custody

Signed, append-only manifests of every copy, move and delete, a record you can hand to an auditor or a court.

Privacy

Your data stays on your disks.

Most backup services start by uploading everything. BackDown starts from the opposite end.

No network by default

The local edition makes no network requests. It needs no account and sends no telemetry.

Never writes to a source

Drives being analysed are mounted read-only, so a flaky enclosure can't corrupt the data you're trying to save.

Never deletes without proof

Nothing is removed unless a copy with a matching hash is already in the journal.

Cloud is opt-in

The optional team dashboard sees metadata only. Your files leave your machines only if you add a cloud backup target.

Planned pricing

Free for what you have. Paid for what it knows.

The core engine will be source-available and free forever. The intelligence, forensic and fleet features are paid, priced per machine rather than per gigabyte.

Community

$0
Free forever, source-available
  • Backup, restore and dedup
  • CLI and local web UI
  • Size analytics
  • Basic recognition
  • Markdown and JSON export

Pro

$8 / month
per machine, or $80 a year
  • Everything in Community, plus
  • Full recognition and provenance
  • Dedup across every source
  • Reclaim simulator
  • Filing suggestions
  • Data viewers
  • Encrypted cloud targets

Team

$24 / month
per machine, 5 machine minimum
  • Everything in Pro, plus
  • Hosted fleet dashboard
  • Shared catalogs and roles
  • Drive health alerts
  • Scheduled attestations

Forensic

$2k / year
per site, from
  • Everything in Team, plus
  • Signed, tamper-evident manifests
  • Chain-of-custody export
  • Air-gapped deployment
  • Priority support

Planned launch pricing in US dollars. It may change before release, and early-access members will hear first.

Get early access.

Linux comes first, as a .deb and an AppImage, followed by macOS and Windows. Leave your email and we'll send one message when it's ready to try.

One email when it's ready. No tracking, and we never share your address.