yandex-disk-rescue: back up all of Yandex Disk with rclone (English guide)

How to copy an entire Yandex Disk, including photos in the “unlimited” photo storage, to a local drive or to Google Drive with rclone, while Yandex restricts downloads and does not serve part of the files.

Everything here was tested on 13 October 2026 on one live account: 1.8 TB of files, 152 GiB (about 30,000 photos) in the unlimited storage, rclone 1.75.2, Windows PowerShell 5.1, bash in WSL (Ubuntu 24.04) and Docker on a VPS. Your account may behave differently.

The scripts print their messages in Russian; the commands and parameters below are all you need.

← README · Russian docs

Contents

Why the usual rclone setup fails

In October 2026, after drone attacks damaged three Yandex data centers (Sasovo, Kaluga, Vladimir), Yandex disabled bulk downloads in the Disk web interface. On the tested account:

So this guide uses the REST backend with a Polygon token, short timeouts so dead files fail fast, copy instead of sync, and MD5 verification.

[!IMPORTANT] Never use rclone sync for this backup. It deletes files in the destination that are missing on Yandex Disk.

1. Install rclone

You need rclone 1.75 or newer.

Windows (PowerShell; reopen PowerShell after installing):

winget install Rclone.Rclone
rclone version

macOS and Linux (distribution packages are often outdated, prefer the official installer on Linux):

brew install rclone                                          # macOS
sudo -v ; curl https://rclone.org/install.sh | sudo bash     # Linux
rclone version

2. Connect Yandex Disk with a Polygon token

Open the Yandex Disk Polygon, sign in to the account you want to back up and click “Получить OAuth-токен” (Get OAuth token). Copy the token. It grants full access to your Disk and is valid for about a year: keep it secret and revoke it when you are done.

The connection script asks for the token (input is hidden), adds a [yandex] remote to the rclone config and prints Total/Used/Free of your Disk.

Windows:

irm https://raw.githubusercontent.com/KenGuru66/yandex-disk-rescue/main/scripts/add-yandex-remote.ps1 -OutFile add-yandex-remote.ps1
powershell -ExecutionPolicy Bypass -File .\add-yandex-remote.ps1

macOS and Linux (an optional argument sets the remote name, default yandex):

git clone https://github.com/KenGuru66/yandex-disk-rescue.git
cd yandex-disk-rescue
./scripts/add-yandex-remote.sh

Manual setup without the script

The rclone yandex backend refuses to start without a refresh_token (unable to get RefreshToken), and a Polygon token has none. Add a placeholder refresh token and a far expiry date, so rclone never tries to refresh it. Append this section to the file shown by rclone config file:

[yandex]
type = yandex
token = {"access_token":"YOUR_POLYGON_TOKEN","token_type":"OAuth","refresh_token":"unused","expiry":"2099-01-01T00:00:00Z"}

Check the connection:

rclone about yandex:

3. Copy to a local drive

The backup script runs rclone copy with fail-fast timeouts, writes a log and a list of files that were not copied (yandex-rescue-<date>-failed.txt) to the current folder. Re-running it copies only what is still missing.

Windows (-Source copies a single folder, -Transfers sets parallel streams, default 16):

irm https://raw.githubusercontent.com/KenGuru66/yandex-disk-rescue/main/scripts/backup.ps1 -OutFile backup.ps1
powershell -ExecutionPolicy Bypass -File .\backup.ps1 -Dest E:\YandexDisk
powershell -ExecutionPolicy Bypass -File .\backup.ps1 -Dest E:\YandexDisk -Source "yandex:Documents" -Transfers 32

macOS and Linux (SOURCE and TRANSFERS environment variables, extra rclone flags after --):

./scripts/backup.sh /Volumes/Backup/YandexDisk                     # macOS
./scripts/backup.sh /media/$USER/Backup/YandexDisk                 # Linux
SOURCE="yandex:Documents" TRANSFERS=32 ./scripts/backup.sh /mnt/backup -- --dry-run

The same copy without scripts

rclone waits 5 minutes and retries 10 times by default, so streams get stuck on files Yandex does not serve. With these flags a dead file fails in 1-2 minutes and a slow but live file still completes:

rclone copy yandex: /mnt/backup/YandexDisk \
  --transfers 16 --checkers 8 \
  --timeout 60s --contimeout 15s --low-level-retries 2 --retries 1 \
  --progress --log-file rescue.log --log-level INFO

List the files that were not copied:

grep " ERROR : " rescue.log

Exit code 5 means some files were not copied: expected while Yandex does not serve them. Run the copy again later.

4. Copy to Google Drive

Files stream from Yandex to Google through your computer or server; no local disk space is needed. Google accepts at most 750 GB of uploads per day, so --drive-stop-on-upload-limit stops the run and the next run continues.

rclone’s shared Google client_id is being retired in 2026, so create your own (rclone docs): a Google Cloud project with the Google Drive API enabled, an External OAuth consent screen with the .../auth/drive scope, and a Desktop app client. Publish the app (“In production”), otherwise the token stops working after 7 days.

rclone config create gdrive drive client_id=YOUR_CLIENT_ID client_secret=YOUR_SECRET scope=drive
rclone about gdrive:

On a server without a browser, create the remote with config_is_local=false and get the token on your computer with rclone authorize "drive" YOUR_CLIENT_ID YOUR_SECRET (remote setup).

Copy with the scripts; for a cloud destination they add --checksum, so every re-run also re-checks MD5:

.\backup.ps1 -Dest gdrive:Yandex-backup

./scripts/backup.sh gdrive:Yandex-backup

Or directly with rclone:

rclone copy yandex: gdrive:Yandex-backup --checksum \
  --transfers 32 --checkers 8 \
  --timeout 60s --contimeout 15s --low-level-retries 2 --retries 1 \
  --drive-stop-on-upload-limit --progress

5. Repeat every day

Each run skips files that are already copied and retries the rest, including files Yandex did not serve yesterday.

Windows Task Scheduler, every day at 03:00:

$a = "-ExecutionPolicy Bypass -WindowStyle Hidden -File `"$PWD\backup.ps1`" -Dest E:\YandexDisk"
Register-ScheduledTask -TaskName "yandex-disk-rescue" -Trigger (New-ScheduledTaskTrigger -Daily -At 3am) `
  -Action (New-ScheduledTaskAction -Execute powershell.exe -Argument $a -WorkingDirectory $PWD)

Remove the task with Unregister-ScheduledTask yandex-disk-rescue -Confirm:$false.

macOS and Linux cron (crontab -e):

17 3 * * * cd $HOME/yandex-disk-rescue && ./scripts/backup.sh /media/backup/YandexDisk >/dev/null 2>&1

6. Verify the copy by MD5

rclone check compares every file by MD5. missing.txt lists files not yet in the copy, differ.txt lists files that differ and should be empty:

rclone check yandex: /media/backup/YandexDisk --one-way --missing-on-dst missing.txt --differ differ.txt

Yandex Disk and Google Drive both store MD5, so checking a Google Drive copy downloads nothing:

rclone check yandex: gdrive:Yandex-backup --one-way --missing-on-dst missing.txt --differ differ.txt

7. Photos in the unlimited storage

Photos from unlimited autoupload are visible only in the “Фото” (Photos) section of the web interface. To back them up, move them to a regular folder first. After the move they count towards your storage quota, so check that you have enough free space. The API shows the size of the unlimited area:

curl -s -H "Authorization: OAuth YOUR_POLYGON_TOKEN" "https://cloud-api.yandex.net/v1/disk/?fields=photounlim_size,total_space,used_space"

Free space is total_space - used_space; it must be at least photounlim_size.

Three ways to get them out:

  1. The Yandex Disk desktop app can copy all photos, including unlimited ones, to your computer (Settings, Autoupload, “Копировать фото и видео на компьютер”). Since 3 June 2026 the desktop app syncs files only with a Yandex 360 subscription. Not tested during the restrictions.
  2. Manually in the web interface: Photos, the “Безлимит” (Unlimited) filter, select up to 500 photos with Shift, Move to a regular folder. Repeat until the section is empty.
  3. Automatically with tools/yd.js, described below.

Move unlimited photos with yd.js

yd.js does what the web interface does: it lists the photounlim area and moves files in batches of up to 40 (Yandex rejects larger batches with Too many items). Requests run inside your own browser tab over the Chrome DevTools Protocol, with your session cookies. On the test account it moved 30,738 files (152 GiB) in about 6 hours; 49 failed with 500 on the Yandex side.

[!WARNING] This is the undocumented internal API of the Yandex Disk web client and can change at any time. The script only moves files, it never deletes them.

You need Node.js 18+ and Chrome or another Chromium browser:

git clone https://github.com/KenGuru66/yandex-disk-rescue.git
cd yandex-disk-rescue/tools
npm install

Create the destination folder on your Disk. The default is Безлимит; to use another folder, set YD_DEST:

export YD_DEST=Unlimited          # PowerShell: $env:YD_DEST = "Unlimited"

Start a separate browser instance with a debugging port and a separate profile, sign in to Yandex in that window and keep the Disk tab open:

# Windows
& "C:\Program Files\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222 --user-data-dir="$env:TEMP\yd-profile" https://disk.yandex.ru/client/photo

# macOS
"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" --remote-debugging-port=9222 --user-data-dir=/tmp/yd-profile https://disk.yandex.ru/client/photo
# Linux (in WSL add --ozone-platform=wayland --disable-gpu)
google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/yd-profile https://disk.yandex.ru/client/photo

List the unlimited area (read-only, 5-15 minutes for 30,000 photos), then move up to 100,000 files in batches of 40:

node yd.js list state
node yd.js move state 100000 40

state is the folder for the file list and the move journal. Running move again resumes and retries only failed files. If Yandex shows a captcha, the script stops: solve it in the browser window and run move again. When finished, close the browser and delete the yd-profile folder: it holds your session cookies.

Tested on Linux (WSL) with Chrome; Windows and macOS commands are the same but were not run.

8. Run on a VPS with Docker

A VPS copies Yandex Disk to the cloud every night without your computer. No disk space is needed on the server, but all traffic passes through it twice. Tested on 1 vCPU and 1 GB RAM with the rclone/rclone:1.75.2 image: 64 streams used about 310 MB of RAM and up to 40% of one core, with speeds up to 33 MB/s.

mkdir -p ~/yandex-backup/config ~/yandex-backup/logs && chmod 700 ~/yandex-backup
cd ~/yandex-backup
curl -fsSLO https://raw.githubusercontent.com/KenGuru66/yandex-disk-rescue/main/vps/run-backup.sh
curl -fsSLO https://raw.githubusercontent.com/KenGuru66/yandex-disk-rescue/main/vps/make-filters.py
chmod 700 run-backup.sh

Put the [yandex] and [gdrive] sections into config/rclone.conf (mode 600; see vps/rclone.conf.example) and check them from the container:

docker run --rm --user $(id -u):$(id -g) -e RCLONE_CONFIG=/config/rclone.conf \
  -v ~/yandex-backup/config:/config rclone/rclone:1.75.2 about yandex:

Start the first run in the background. Variables: DEST (required), SOURCE (default yandex:), TRANSFERS (32), BWLIMIT (off or e.g. 20M), CPUS and MEMORY (container limits, 0.85 and 550m):

DEST=gdrive:Yandex-backup setsid nohup ./run-backup.sh > logs/run.out 2>&1 < /dev/null &
tail -f "$(ls -t logs/backup-*.log | head -1)"

Nightly run with cron; flock inside the script prevents overlapping runs, logs older than 30 days are deleted:

23 2 * * * DEST=gdrive:Yandex-backup /home/USER/yandex-backup/run-backup.sh >> /home/USER/yandex-backup/logs/cron.out 2>&1

Save important folders first, skip dead files

During the first upload, dead files waste time on timeouts. After the first pass, build a filter from the logs and copy the important folders first. Filter paths are relative to the SOURCE of the logged runs:

python3 make-filters.py logs "Documents" "Photos/2024"                        # filter-all.txt, filter-1.txt, filter-2.txt
DEST=gdrive:Yandex-backup ./run-backup.sh --filter-from /logs/filter-1.txt    # only Documents, without dead files
DEST=gdrive:Yandex-backup ./run-backup.sh --filter-from /logs/filter-all.txt  # everything else, without dead files

Keep the regular nightly run without a filter: it retries every file and picks up those Yandex brings back.

9. Errors

In the log Meaning Fix
410 GoneError on any folder the remote was created by rclone config with rclone’s own app reconnect with a Polygon token (step 2)
unable to get RefreshToken no refresh_token in the [yandex] section add "refresh_token":"unused" (step 2)
401 Unauthorized the Polygon token expired or was revoked get a new token and replace access_token
timeout awaiting response headers, 500 Internal Server Error the file sits on a Yandex storage node that does not answer now nothing; run the copy again later
failed to open source object: 503 WebDAV is overloaded use the REST yandex remote, not WebDAV
Google 403 User rate limit exceeded more than 750 GB uploaded today continue tomorrow
Too many items in yd.js batch larger than 40 node yd.js move state 100000 40
exit code 5 some files were not copied expected now; see *-failed.txt and run again later

10. Revoke the token when you are done

The Polygon token gives full access to your Disk for about a year.

rclone config delete yandex

Then, in Yandex ID, open the list of apps with access to your account and disconnect the Yandex Disk Polygon app. If you cannot find it, change your password and sign out on all devices.