Cross-platform cloud backup with serious deduplication tech, honestly reviewed. Including the parts the website doesn’t mention.

TL;DR

What it is: A cross-platform backup tool built on “Lock-Free Deduplication” — a technically novel approach that lets multiple machines share a single cloud storage without stepping on each other [README][5].
Who it’s for: Sysadmins and homelab operators running multiple servers who want aggressive deduplication across machines to a single cloud destination. Less suited to non-technical users who expect a polished GUI.
Cost: CLI is free for personal use; commercial CLI license costs $50/machine/year. The Web GUI is a separate paid product — and the forum community has opinions about whether that price is justified [1][website].
Key strength: Cross-computer deduplication is genuinely unique. No other backup tool in its class lets Machine A and Machine B deduplicate against each other’s chunks without direct communication [README][5].
Key weakness: The Web GUI is widely described as “half-finished” and “broken in all kinds of ways” by paying customers [1]. The CLI is capable but not beginner-friendly. And the license is neither cleanly open-source nor straightforwardly commercial — it sits in an uncomfortable middle ground that frustrates developers who want a purely free solution [2].

What is Duplicacy

Duplicacy is a backup tool built around one core innovation: Lock-Free Deduplication. The idea is that chunk deduplication — saving only the changed blocks of your files — normally requires a centralized database that every backup client locks before writing. That database becomes a bottleneck and a single point of failure. Duplicacy eliminates it by storing each chunk as an independent file named by its own hash, so any client can look up whether a chunk already exists using a basic filesystem operation, without holding a lock [README].

The practical payoff is that you can point ten different servers at the same S3 bucket, and they’ll all deduplicate against each other without any coordination layer. A 50MB shared library that lives on all ten machines gets stored once [README][5]. This is the feature that separates Duplicacy from Borg, restic, Kopia, and most other deduplication tools, where cross-machine deduplication either doesn’t exist or requires a shared repository that serializes writes.

The project sits at 5,634 GitHub stars. The command-line version is the core product — source code is published on GitHub, free for personal use, with commercial licenses sold separately. On top of that sits a Web GUI (the “Web Edition”), sold as a separate paid product, which is where most of the community friction originates.

A note on the license: the GitHub repository shows “NOASSERTION” for the license type. This is a source-available model, not an OSI-approved open-source license. You can read the code, but redistribution and commercial use are restricted. For founders and teams who want a tool they can self-host, embed, or hand off without legal worry, this matters [2].

Why People Choose It

The pitch that lands for Duplicacy is specific: multiple machines, one storage backend, maximum efficiency. If you’re backing up a homelab with five nodes to Backblaze B2, the cross-computer deduplication means your OS base layer, Docker images, and shared application files are stored once rather than five times. That’s not a marginal saving — on a large enough setup, it can cut your storage bill in half or more [README][5].

The MangoHost comparison [5] between Duplicacy, restic, and Borg puts this plainly: Duplicacy uses global deduplication across all clients sharing the same storage. Restic and Borg deduplicate within a single repository — if you use separate repos per machine, you get no cross-machine savings at all. For homelab operators or small VPS fleets, Duplicacy’s model is architecturally superior.

Performance is also a genuine strength. The README links to an academic benchmarking experiment backing up the Linux kernel source against Duplicacy, restic, Attic, and duplicity — Duplicacy comes out fastest by a notable margin. The cloud storage comparison in the README shows consistent throughput across S3, Backblaze, Azure, and GCS [README].

The storage backend list is comprehensive: local disk, SFTP, S3 (and any S3-compatible), Backblaze B2, Google Drive, Azure, Dropbox, OneDrive, Google Cloud Storage, DigitalOcean Spaces, Wasabi, WebDAV, OpenStack Swift, and more [README][website]. Eric Cheng, writing in 2022 [2], noted the Google Drive support as a specific draw before he ultimately moved to Kopia.

What pushes people away is a different story. The same Eric Cheng piece describes Duplicacy as “straightforward to use” but ultimately rejected it because he “wasn’t comfortable paying licensing fees to solve a problem that ought to have a free and open-source solution” [2]. That’s the friction point that appears repeatedly: the technology is well-regarded, the licensing philosophy is not.

Features

Core backup engine:

Lock-Free Deduplication — no centralized chunk database, no locking, no single point of failure [README]
Cross-computer deduplication to a shared storage backend [README][5]
Client-side encryption (AES-GCM) before data leaves your machine [README][5]
Incremental backups — only changed chunks are uploaded
Snapshot-based model — each backup is a complete addressable snapshot, not a delta chain [README]
Flexible retention policies via prune command (-keep 0:180 -keep 7:30 -keep 1:7 syntax) [5]
RSA asymmetric encryption support for stronger key separation [README]
Erasure coding for resilient multi-destination protection [README]
Versioned snapshots — restore any previous revision [README]
File-level and directory-level restore [README]
diff, history, cat, check commands for inspecting and auditing backup state [README]
Filters / include-exclude patterns for precise scope control [README]

Storage backends (16+): S3, Backblaze B2, Google Drive, Google Cloud Storage, Azure, Dropbox, OneDrive, DigitalOcean Spaces, Wasabi, SFTP, local disk, WebDAV (beta), pCloud, Box.com, OpenStack Swift, File Fabric [README]

Interfaces:

CLI: single binary, scriptable, cross-platform (Linux, macOS, Windows) [README][5]
Web GUI (Web Edition): separate paid product — dashboard with charts, scheduled jobs, visual storage management [website]
No native mobile app

What’s missing or limited:

No built-in scheduling in the CLI — you add your own cron or task scheduler
No built-in monitoring or alerting — the forum wiki has guides for healthchecks.io and similar [1]
The Web GUI lacks features that users expect for the price — basic configuration gaps have been reported by paying customers [1]

Pricing: SaaS vs Self-Hosted Math

This is where Duplicacy’s model gets complicated and where the community friction makes sense.

CLI pricing:

Personal use: free
Commercial use: $50 per machine per year [website]

“Commercial” is defined as use in a business context. If you’re backing up servers for a company, that’s commercial. For a freelancer with a VPS, the line is blurry.

Web GUI (Web Edition):

One-time purchase — the forum post from a frustrated user [1] mentions “$20 for some half finished web version,” which matches the pricing tier for a single-machine GUI license at time of writing
The website describes volume discounts for IT service providers [website]

Comparison with alternatives:

restic: 100% free, BSD-2 license, no GUI but no licensing confusion [5]
Borg: 100% free, BSD license, limited cloud support [5]
Kopia: 100% free, Apache license, comparable deduplication, actively maintained — the tool Eric Cheng [2] landed on after leaving Duplicacy
Backblaze B2 (example storage): ~$6/TB/month. With aggressive cross-machine deduplication, a 5-node homelab that would otherwise consume 5TB might only consume 1TB — saving ~$24/month on storage. That math can justify the CLI license cost if the deduplication ratios are high enough.

For a solo homelab user: the personal CLI license is free, and the math is clearly favorable if you’re paying for cloud storage and the deduplication actually lands. For a business running five servers, $250/year ($50 × 5) for the CLI licenses is modest compared to what you’d pay in cloud storage without deduplication. The Web GUI value proposition is harder to defend given community feedback about its quality.

Deployment Reality Check

The CLI setup is straightforward for anyone comfortable with a terminal. The mangohost.net guide [5] distills it to four commands:

curl -L https://github.com/gilbertchen/duplicacy/releases/latest/download/duplicacy_linux_x64 -o /usr/local/bin/duplicacy
chmod +x /usr/local/bin/duplicacy
duplicacy init my-backup s3://my-bucket/path
duplicacy backup

Single binary, no dependencies, runs on Linux, macOS, and Windows. Credentials for each storage backend are set via duplicacy set -key ... -value ... or environment variables. The wiki covers every storage backend with specific setup steps [README].

Where it gets harder:

Scheduling: The CLI has no scheduler. You write a cron job or a systemd timer yourself. The forum wiki has guides, but this is a step that catches non-technical users [1].

Web GUI: The Web Edition installs separately and exposes a browser-based dashboard. The forum community’s verdict, including from a long-time user with apparent technical depth, is unambiguous: “The WebUI is a toy, that needs to be scrapped and outsourced” and “horrible, unusable, broken in all kinds of ways” [1]. The same user describes the CLI engine as “rather great” — the criticism is specifically about the GUI, not the backup core.

macOS specific: The forum [1] surfaces macOS as a problematic environment — backing up all users or user-specific mount points is described as difficult to configure without terminal access.

Restore testing: Any backup tool needs to be verified by actually restoring. The check command validates chunk integrity, and the restore command works cleanly per the benchmark write-ups [5]. But always test restore before relying on it in production — this is table stakes with any backup system.

Maintenance concerns: A forum thread from 2023 [1] titled “Is Duplicacy no longer maintained?” raises concerns about low-priority CLI updates and slow bug fixes (a Dropbox authentication break reportedly took a long time to resolve, and an interrupted prune bug affecting subsequent checks was still open at time of writing). The academic paper on lock-free deduplication was published in IEEE Transactions on Cloud Computing [README], which suggests the core technology is well-documented — but research-grade documentation and production-grade support velocity are different things.

Realistic setup time for the CLI on a fresh Linux server: 15–30 minutes to first backup. For the Web GUI on a clean install: 1–2 hours including configuration and troubleshooting the documented rough edges.

Pros and Cons

Pros

Cross-computer deduplication is genuinely unique. No other widely-used backup tool in this class lets multiple machines deduplicate against each other to a shared cloud backend without direct communication or a centralized coordination server [README][5]. If you run multiple servers, this is a real differentiator.
Database-less architecture reduces failure modes. No central chunk database means no database corruption, no single file that can go corrupt and lose your backup index [README]. Each chunk is independently addressable.
Broad cloud storage support. 16+ backends including every major cloud provider and WebDAV. Wasabi and Backblaze B2 support are explicit, which matters for cost-conscious self-hosters [README].
Strong encryption built in. AES-GCM client-side encryption with optional RSA asymmetric key support and erasure coding [README][5]. Data is encrypted before it leaves your machine.
Single binary CLI. No runtime dependencies, runs everywhere, scriptable [README][5].
Benchmarks well against restic, Borg, duplicity. Published benchmarks on the Linux kernel codebase show Duplicacy ahead on speed [README].
IEEE-peer-reviewed algorithm. The lock-free deduplication design has been published in academic literature — that’s more rigor than most backup tools can claim [README].
Personal use is free. For home users and hobbyists, the CLI costs nothing.

Cons

The Web GUI is widely panned. Paying customers in the official forum describe it as “horrible, unusable, broken in all kinds of ways” [1]. The CLI engine gets praise; the GUI gets complaints. If you need a working GUI, look elsewhere.
Not truly open source. The license is source-available, not OSI-certified open source. You can read the code but redistribution and commercial use require a paid license [2]. For founders who want a tool they can embed, fork, or hand off, this is a dealbreaker.
Maintenance pace concerns. Forum users in 2023 raised questions about update frequency and slow bug resolution [1]. The Dropbox backend breakage and an unresolved prune bug are cited specifically.
No built-in scheduling. Cron or systemd required — not a problem for sysadmins, a barrier for everyone else.
macOS rough edges. Backing up across users or certain macOS mount points requires terminal configuration that isn’t well-documented [1].
CLI-first means non-technical users are underserved. The target audience for the Web GUI is presumably non-technical users, but that product has the worst reviews [1].
Competing free alternatives have caught up. Kopia and restic are fully free, actively maintained, and for most single-server use cases, the cross-machine deduplication advantage doesn’t apply [2][5].

Who Should Use This / Who Shouldn’t

Use Duplicacy if:

You’re managing multiple servers or nodes backing up to the same cloud storage, and you want cross-machine deduplication to cut storage costs.
You’re comfortable with the CLI and will run it via cron — the Web GUI is optional for you.
You’re an individual or personal user where the free personal license applies and storage backend breadth matters.
You specifically need Backblaze B2, Wasabi, or one of the less common backends that restic doesn’t natively support.

Skip it (pick restic instead) if:

You’re backing up a single machine and don’t need cross-computer deduplication. Restic is fully free (BSD-2), single binary, actively maintained, and covers S3, B2, Azure, and SFTP natively [5].
You want a tool with zero licensing ambiguity. Restic is unambiguously open source — no personal-vs-commercial gray area.

Skip it (pick Kopia instead) if:

You want a fully free, Apache-licensed alternative with a working Web UI and comparable deduplication. Eric Cheng [2] landed on Kopia after trying Duplicacy and describes it as “to backups as WireGuard is to VPNs.”
You need a GUI that actually works without forum complaints about it being “unusable.”

Skip it (pick Borg instead) if:

You’re backing up to SSH/SFTP targets and don’t need cloud backends. Borg is the fastest option for local and SSH destinations [5].

Don’t use the Web GUI if:

You’re a paying customer expecting a polished product. The community consensus is that it doesn’t meet that bar [1].

Alternatives Worth Considering

restic — Fully BSD-licensed, single binary, strong S3/B2/Azure support, no GUI. The most direct free alternative to the Duplicacy CLI. Per-repository deduplication only [5].
Kopia — Apache-licensed, free, has a working Web UI, comparable deduplication, active development. Recommended by users who left Duplicacy over licensing [2].
Borg — BSD-licensed, excellent for SSH and local targets, per-repository deduplication. Limited native cloud support but pairs with rclone [5].
Duplicati — Free, has a Web UI, broad cloud support. Has a reputation for database corruption on Linux [4] — the Duplicati forum developers are open about this being an ongoing area of work.
Rclone — Not a backup tool itself but pairs with Borg for cloud sync, and many users combine it with the tools above for cloud targets that aren’t natively supported.

Bottom Line

Duplicacy’s lock-free deduplication is a genuine technical achievement — the cross-computer deduplication model is unique in the open-source backup space and produces real savings for anyone running multiple machines against a single cloud storage backend. The CLI engine is well-reviewed, the benchmark numbers are strong, and the storage backend support is comprehensive.

But the licensing story undermines it. The source-available-not-open-source model sits in an uncomfortable middle: not free enough for developers who want a clean OSI license, not polished enough for users who expect a working GUI for their money. The Web Edition has visible quality problems that paying customers complain about openly in the project’s own forum [1]. And the maintenance pace has raised questions that haven’t been fully answered.

For a solo sysadmin backing up five homelab nodes to Backblaze B2, the free personal CLI license and the cross-machine deduplication math can absolutely make sense — just don’t expect the Web GUI to save you. For everyone else, restic or Kopia give you comparable backup quality without the licensing friction and without the GUI disappointment.

Sources

Duplicacy Forum — “Dont buy this scam” (community discussion including developer and power-user responses on GUI quality and maintenance). https://forum.duplicacy.com/t/dont-buy-this-scam/7617
Eric Cheng, chengeric.com — “Backups: the good, the bad, and the ugly” (personal migration story comparing Duplicity, Duplicati, Duplicacy, Timeshift, and Kopia). https://www.chengeric.com/backups/
SourceForge — “Duplicacy Reviews in 2026” (product listing with alternatives). https://sourceforge.net/software/product/Duplicacy/
Duplicati Forum — “Why the negative reviews?” (community comparison of Duplicati, Kopia, restic, and Duplicacy; developer responses on architecture). https://forum.duplicati.com/t/why-the-negative-reviews/19693
MangoHost Blog — “Duplicacy vs Restic vs Borg: Which Backup Tool is Right in 2025?” (feature comparison table, setup examples, license analysis). https://mangohost.net/blog/duplicacy-vs-restic-vs-borg-which-backup-tool-is-right-in-2025/