This is the Linux edition of the S3 GUI client overview. The Mac edition covers Homebrew and mounting in Finder.
Linux is where the S3 GUIs stop
Look at what the category actually ships. S3 Browser is Windows-only. Cyberduck is free, GPL and has existed since 2002, and it builds for macOS and Windows — there is no Linux version, and as of September 2026 Flathub carries FileZilla but not Cyberduck. CloudMounter, Transmit and ForkLift are Mac software. What is left on a Linux desktop is rclone, s3cmd and the AWS CLI: all excellent, all typed.
That is fine until the task is visual — spotting which of four similarly named exports is the 900 MB one, dragging two out, checking that last night's backup job wrote anything. AnyStorage is a desktop app for that layer, and Linux is not a port: one Tauri v2 release ships for macOS, Windows and Linux on the same day.
What you download
Release 0.2.24 went out on 23 September 2026 with two Linux assets, both x86-64:
- amd64.AppImage, about 88 MB. Mark it executable with chmod +x and run it. Nothing is installed system-wide, which makes it the easy option on Fedora, Arch, openSUSE or anything else that is not Debian-shaped.
- amd64.deb, about 9.1 MB. On Debian and Ubuntu install it with sudo apt install ./AnyStorage_0.2.24_amd64.deb so dependencies resolve.
The supported floor is Ubuntu 20.04. Both files sit on the release page at github.com/Kilerd/anystorage-release/releases beside the macOS .dmg and the Windows installers, and the app checks a latest.json manifest published with every release, so a new version announces itself.
Two gaps: no Flatpak, Snap or AUR package yet, and no arm64 Linux build — the Linux assets are amd64 only.
Connecting your first bucket
Amazon S3 asks for four things: an access key ID, a secret access key, a region, and the bucket you want to see. The app inherits exactly the permissions of the key you hand it, so hand it an IAM user limited to one bucket and one prefix rather than your root keys.
S3-compatible endpoints need two extra settings, and they are the two people get wrong:
- A custom endpoint URL, so the client talks to your server instead of amazonaws.com. This is what makes MinIO, Wasabi and Ceph RGW work.
- Path-style addressing, which MinIO and several other implementations require instead of virtual-host-style bucket URLs. Leave it off for Amazon; turn it on when a self-hosted endpoint answers with nothing but signature or DNS errors.
Cloudflare R2, Backblaze B2, Google Cloud Storage, Azure Blob, Alibaba Cloud OSS and Tencent Cloud COS have their own connection types rather than being S3 with extra steps, out of more than 20 types in total — WebDAV, FTP, SFTP, Google Drive, OneDrive, Dropbox and local folders among them. Credentials are stored encrypted on the machine and there is no account to create.
Browsing and moving files
Buckets open as folders, prefixes behave like directories, and tabs keep several locations open at once, so a project prefix, a backup path and a delivery folder stay where you left them. Uploads and downloads land in one transfer area with progress and history.
Pro adds the parts that turn browsing into work: search across a connection, in-app preview of images, text and audio, bulk operations, drag-and-drop upload, and presigned links — a time-limited URL is easier to hand over than a new user on an internal MinIO server.
Mounting a bucket as a folder, without FUSE
Some programs take nothing but a path. On Linux the usual answers are s3fs-fuse and rclone mount: both work, both put a FUSE process between your editor and the bucket, and both mean one mount per remote.
AnyStorage goes around it. The app talks S3 over the ordinary API, then re-serves everything currently connected through a WebDAV server running on your own machine — no kernel module, no fusermount, no root.
Start it from Services in the sidebar: the WebDAV tab, then Start. It listens on http://127.0.0.1:3211 by default, the port is editable anywhere in 1–65535, and HTTP Basic auth is always on: username anystorage, password 24 random characters generated on first use. Then mount it with what your desktop already has:
- GNOME Files (Nautilus): Other Locations, then dav://127.0.0.1:3211 in the Connect to Server field.
- KDE Dolphin: webdav://127.0.0.1:3211 in the location bar.
- No desktop at all: davfs2, with sudo mount -t davfs http://127.0.0.1:3211 /mnt/anystorage, which asks for the same credentials unless they sit in /etc/davfs2/secrets.
Every connection currently open in the app appears as a folder at that root, named after the connection, so a bucket, an SFTP host and a Drive account become siblings under one mount. Reads stream with HTTP range support: pulling the playhead into the middle of a large video in mpv fetches the bytes around that point.
The limits belong in the same breath. On the free tier the WebDAV server is read-only. There is no TLS in the current version, which is why it binds to loopback; a LAN mode binding 0.0.0.0 exists and the app warns you, because Basic auth over plain HTTP is credentials in the clear. After an app restart a connection appears under the WebDAV root only once you open it in the app, because connections connect lazily. You also cannot copy between two connection folders through the mount — cross-storage moves belong in the transfer view.
Against the tools you already have
rclone is the strongest thing on this list and we will not pretend otherwise: around 60 backends, scriptable, free, and on Linux its FUSE dependency costs nothing because FUSE is already in the kernel. If your work is sync jobs and cron, stay there. The difference is shape: rclone mount is one process per remote plus a --vfs-cache-mode decision before many applications will write to it, where this is one endpoint and one mount point for every connection.
s3cmd and aws s3 cp are right for a copy you can describe in advance, and wrong for finding a file whose name you only half remember, three prefixes deep.
Cyberduck and S3 Browser are the two names people arrive with, and neither runs here — macOS and Windows for the first, Windows only for the second. That is the real state of the category on Linux, and it is why "s3 browser linux" has no comfortable answer.
What the free tier covers
The free tier allows two storage connections and basic file operations, and it permits commercial use: no personal-use-only clause, no trial timer. WebDAV is read-only there, and support is community-based. Pro lifts the connection limit, opens all 20-plus connection types, makes WebDAV read-write, and adds search, preview, bulk operations, drag-and-drop upload and priority email support.
Questions
Does it work on Ubuntu 22.04 or 24.04?
Yes — 20.04 is the floor, so both are above it. Install the .deb on Debian and Ubuntu; elsewhere use the AppImage.
Is there a Flatpak, Snap or AUR package?
Not yet. AppImage and .deb are the only Linux packages today; macOS has a Homebrew cask (brew install --cask kilerd/tap/anystorage).
Do I need FUSE, macFUSE or a kernel module?
No. Browsing happens in the app over the S3 API, and the optional folder-style mount uses the local WebDAV server plus your file manager's existing WebDAV client. Nothing lands in the kernel, and nothing needs root except davfs2.
Can it connect to MinIO, Cloudflare R2, Wasabi or Ceph?
MinIO, Wasabi and Ceph RGW go through the S3 connection type with a custom endpoint URL and path-style addressing. R2 has its own connection type, as do Backblaze B2, Google Cloud Storage, Azure Blob, OSS and COS.
Is there a command line or a headless mode?
No. This is a desktop GUI; on a display-less server rclone or the AWS CLI is still the right answer.