Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
upstream: github.com/duplicati/duplicati (opens in new tab) · github.com/kopia/kopia (opens in new tab)
Pick Duplicati
If you want the simplest path to a scheduled, encrypted backup with a web UI, especially to a consumer cloud drive or a less common provider, and are fine with a local database per job that is rebuilt from the destination if lost.
Pick Kopia
If you want policies, automatic maintenance, an Apache-2.0 licence throughout, a repository server for several machines and the lighter idle footprint we measured (~7 MB against ~103 MB).
Duplicati and Kopia are the two backup tools in our catalog that ship a web UI, and for someone who does not want to write cron lines, that is the whole appeal. Both encrypt on the client, both split data into blocks and upload only what is new, and both can target S3, Backblaze B2, SFTP and WebDAV. They differ in language and design: Duplicati is a C# client with a local SQLite database per backup job and a long list of cloud destinations; Kopia is a Go tool with a repository, policies and a server mode. If you are also weighing the command-line option, restic vs Kopia covers it.
Duplicati's README calls it "a free, open-source backup client that securely stores encrypted, incremental, and compressed backups". You define a backup job in the web UI (source folders, destination, schedule, retention, passphrase), and the "built-in scheduler" runs it. After each upload, "a few random volumes are downloaded to ensure the backup data is available". Per upstream, the backup process reads a block at a time, ignores blocks that already exist, packs new ones into volumes that are compressed, encrypted and uploaded, and finishes by uploading a file list that "explains how to recreate all the processed files".
Kopia's features page pitches "end-to-end encryption, with user-controlled encryption keys", and says "Multiple copies of the same file will be stored once." It "creates snapshots of the files and directories you designate" into a repository, with "an unlimited number of policies for each repository". It also lists Reed-Solomon error correction, built-in functions "to verify the consistency/validity of your backed up files", and "a powerful official graphical user interface".
Both end up in the same place: encrypted, deduplicated backups you can browse and restore from a browser.
Duplicati's built-in encryption is "an open-source C# port" of the AES Crypt
format. Each volume gets its own random key, and an HMAC signature detects
modification. Encryption is a per-job choice: the setup page lets you opt out,
and upstream asks that you then "control the storage destination". GPG is the alternative; it needs the GPG binaries and defaults to
symmetric encryption with CAST-128 and your passphrase, with custom arguments
available for asymmetric setups. In Docker you can also encrypt Duplicati's own
settings database with SETTINGS_ENCRYPTION_KEY, which upstream says "is not the
password used to connect to the UI".
Kopia encrypts the repository end to end with keys you control, using AES-256-GCM by default or ChaCha20-Poly1305, picked when you create the repository, and its features page states that it "does not allow creating unencrypted backups". Our install passes it as an environment variable so the container can reopen the repository after a restart, which means you must keep that value somewhere other than the server you are backing up.
This is the structural difference. Duplicati keeps a local database per backup,
and the docs say that "If the local database is missing, it is recreated from
the remote storage" when you run the repair command. The database is a cache
you can rebuild; upstream warns that "it may take a while to recreate the
database", but that restores are possible without it. In the Docker image the configuration lives in /data, so mount a
volume there.
Kopia keeps the state in the repository itself. Maintenance is automatic: quick maintenance "will execute approximately every hour" and full maintenance "will execute every 24 hours", the latter doing snapshot garbage collection and compacting packs. Only the maintenance owner runs it; "Other repository users will not attempt to run maintenance automatically."
Duplicati compresses as part of the volume pipeline and offers "Smart backup retention", which "keeps one daily backup and then gradually fewer versions going back in time". With retention rules active, it removes old file lists after the backup and may consolidate remote volumes to reclaim space.
Kopia stores retention, schedule, ignore rules and compression in policies, set per path or globally. Compression is off by default; its FAQ names compression and parallelism as "the two big culprits of memory usage in Kopia when creating snapshots", and a setting applies only to data uploaded afterwards. Turn it on in the global policy on day one.
Duplicati's destination overview is broad: File, S3-compatible, FTP, SFTP, WebDAV, OpenStack, Rclone, SMB, Backblaze B2, Amazon S3, Azure Blob Storage, Google Cloud Storage, Dropbox, Google Drive, OneDrive, Box.com, pCloud, Storj and more. Some are marked deprecated or untested. If your backup target is a consumer cloud drive, this list is why people pick Duplicati.
Kopia supports "Amazon S3 and any cloud storage that is compatible with S3", plus B2, Azure, Google Cloud, SFTP, WebDAV, rclone, a local or network disk, and its own repository server. Anything else goes through rclone, which upstream marks as experimental.
Kopia has a Repository Server for one shared repository with per-user access, described on restic vs Kopia. Duplicati's central management is the Duplicati Console, and the docs note that "The Agent mode is only available on the Enterprise plan"; the Console is optional, and the first-start page "offers to connect Duplicati to the Duplicati Console".
On licences, "The Duplicati core is licensed under the MIT license", while "The
contents of the proprietary/ folder are covered by the Duplicati Inc Software
license". Kopia is Apache-2.0 throughout.
Neither project publishes a minimum RAM figure; the 512 MB floor on both of our pages is our own estimate. In our install check (GCP e2-standard-2, Ubuntu 26.04), idle at first boot with the default install snippet and nothing backed up, Duplicati used ~103 MB of RAM and ~683 MB of disk (October 2026), and Kopia used ~7 MB of RAM and ~300 MB of disk (September 2026). Those are idle numbers, not a backup in progress: scanning and uploading a large tree will use more on either, and Kopia's own FAQ says so about compression.
Duplicati is the easiest: one container with the UI on port 8200, a web password
set through DUPLICATI__WEBSERVICE_PASSWORD, and your host paths mounted into
it (we mount /home). Without a password, the first-launch password is random and
you sign in using a link printed in the Docker logs. We rate it 1 out of 5.
Kopia is 2 out of 5: a server-mode container on loopback with a UI password
and a repository password, reached over an SSH tunnel, then you create the
repository (S3, B2, SFTP) in the UI. Its --insecure flag means no TLS, so keep
it on loopback.
Choose Duplicati if you want the simplest path to a scheduled, encrypted backup with a web UI, especially to a consumer cloud drive or an unusual provider from its long destination list, and you are comfortable with a local database per job that can be rebuilt from the destination.
Choose Kopia if you want policies, automatic maintenance and a repository that carries its own state, an Apache-2.0 licence throughout, a repository server for several machines, and the lighter idle footprint we measured. Switch compression on and expect a slightly longer first setup.
Both installs are verified on Ubuntu 26.04; the pinned compose files are on the
Duplicati and Kopia pages. Either one
fits the smallest plan most providers sell. Mount what you want backed up
read-only where you can (ours mounts /srv read-only for Kopia), keep the
passphrase or repository password off the server, and test a restore before you
trust it. Neither tool replaces a database dump: dump first, then back up the dump.
No. Duplicati is MIT, Kopia is Apache-2.0.
Duplicati for the quickest setup: one container, a web UI on port 8200, a scheduler and a long destination list that includes consumer cloud drives. Kopia for policies, automatic maintenance, Apache-2.0 licensing and a lighter idle footprint (~7 MB against ~103 MB in our first-boot check, which is idle, not mid-backup).
Duplicati's docs say that if the local database is missing, the repair command recreates it from the remote storage. The backups themselves live at the destination, so nothing is lost; upstream warns a recreate may take a while, but restores are possible without the database.
No. Compression is off by default and set through policies, and a new setting applies only to data uploaded afterwards. Kopia's FAQ names compression and parallelism as the main causes of memory use when creating snapshots, so enable it in the global policy early and watch memory on a small VPS.
The README says the core is licensed under the MIT license, while the contents of the proprietary/ folder are covered by the Duplicati Inc Software license. Separately, Duplicati's docs say Agent mode, used with the optional Duplicati Console, is only available on the Enterprise plan. Kopia is Apache-2.0 throughout.
Paid link — we earn a commission if you shop through it.
The scriptable backup binary vs. the policy engine with a web UI.
A whiteboard-and-docs workspace on a plain Compose file vs. native apps on a commercial server.
A full identity platform vs. a passkey-only sign-in box.