Proton Drive on Linux: Status as of July 2026
The official Linux client has been announced but is not yet available. On servers, Proton Drive can currently be mounted with Rclone; the new SDK indicates the technical direction. What is still missing is machine access limited to individual folders or tasks.
Proton Drive has offered dedicated sync clients for Windows and macOS since 2023. On Linux, there is still only the web interface, community tools, and an official SDK in preview. The situation is even more difficult on a server, where neither desktop sync nor interactive sign-in is a good fit.
This overview describes the status as of July 2026. In addition to the published roadmaps, it is based on a practical test of the Rclone backend as document storage for Paperless-ngx.
The Linux client has been announced, but has no date yet
In June 2026, Proton explicitly confirmed for the first time that a Linux client is being developed. It is being built on the new unified SDK and is intended to use the same technical foundation as the Windows and macOS applications. There is no release date or public beta yet.
Important context: this will be a desktop sync client. It solves the problem for the desktop. For server applications, however, a sync client is the wrong tool, because a service needs to read files directly from Proton Drive and write files to it. A sync client maintains a complete local copy, precisely what should be avoided when storage is limited.
Today, Rclone does the practical work
On Linux, Rclone with its protondrive backend is currently the most versatile tool. It can copy and synchronize files and, as the only available solution, make Proton Drive available like a local directory via a FUSE mount. Two limitations are important:
It is beta software built on a reconstructed API. Proton does not publicly document its Drive API; the backend is based on reverse engineering. In testing, it worked reliably, but throttled rapid sequences of calls with inconsistent directory listings.
For unattended operation, Rclone asks for the TOTP key. The configuration wizard labels the field otp_secret_key. This means the permanent key from the 2FA setup, not the six-digit code currently displayed by an authenticator app. Rclone stores this value obfuscated and generates a valid TOTP code from it for each sign-in.
Anyone who accidentally enters a current one-time code can complete the initial sign-in. However, the next reauthentication fails with error 8002 because Rclone cannot use the same code again.
This keeps the account protected against an isolated stolen password. However, a compromised server exposes both the password and TOTP key. A dedicated Proton account is therefore recommended for automated access.
How such a mount behaves in Docker environments, including two undocumented issues, is covered in the separate article on Rclone in containers.
The official SDK shows where development is headed
At the same time, Proton is migrating its applications to an official SDK for JavaScript and C#, with bindings for Swift and Kotlin. The public repository also includes a command-line tool. Its sign-in model is cleaner than that of the Rclone backend:
auth loginopens the browser; sign-in proceeds normally including two-factor authentication- the session is stored in the operating system’s key store (Keychain, Credential Manager, libsecret), and the SDK renews it itself
- afterward: list files with machine-readable JSON output, upload files, and check shares
The password and TOTP key therefore do not need to be stored in a configuration file. Three limitations remain for server use: the CLI cannot mount a file system, sign-in opens a browser, and Proton does not yet consider the SDK production-ready for third-party applications. Release is planned for late 2026 to early 2027.
The real gap: machine access
The core problem lies one level deeper than the client or SDK: Proton has no machine access. No app password, no service account, no token with limited scope. Every automation, whether a backup script, server mount, or CI job, must work with the account’s full credentials.
For comparison, access key pairs are standard for S3-compatible storage, revocable and restrictable to buckets or prefixes. Google and Microsoft offer app passwords and service accounts. With Proton, however, it is all or nothing: anyone who wants to give a server access to one folder gives it the entire account.
With an end-to-end encrypted service, this is more difficult than with S3 because limited access would also have to mean limited key material. However, the SDK sessions show that Proton can handle such constructs. A session is already a derived, revocable access credential. An official “machine token for exactly this folder, read-only” would be the single biggest advancement for server use, far ahead of any client.
Recommendation by use case
| Use case | Status as of July 2026 |
|---|---|
| Desktop sync on Linux | Wait for the announced client; until then, use Rclone sync or the web interface |
| Server backup (uploading files) | Rclone with copy or sync; works, but account for beta status |
| File system mount for services | Rclone with mount, a stored TOTP key, and a dedicated account; the only field-tested approach |
| Script automation with clean sign-in | Keep an eye on the SDK CLI; still too early for production |
On the Linux desktop, you can wait for the announced client or use Rclone for the time being. On servers, Rclone remains the only practical mount solution. However, a working workaround will only become a robust platform when Proton offers limited machine access and an officially supported mount.
Comments
Comments are loaded from GitHub / Giscus.