Self-Hosted Sync Server

Advanced users who cannot or do not wish to use AnkiWeb can use a self-hosted sync server instead.

Things to be aware of:

• This is an advanced feature, targeted at users who are comfortable with networking and the command line. If you use this, the expectation is you can resolve any setup/network/firewall issues you run into yourself, and use of this is entirely at your own risk.
• Newer clients may depend on changes to the sync protocol, so syncing may stop working if you update your Anki clients without also updating the server.
• Third-party sync servers also exist. No testing is done against them, and they tend to take time to catch up when the sync protocol changes, so they are not recommended.
• The messages inside Anki will use the term "AnkiWeb" even if a custom server has been configured, (e.g. "Cannot connect to AnkiWeb" when your server is down).

Installing/Running

There are various ways you can install and run the server. You can use either:

• the sync server bundled with the desktop version of Anki
• a separate minimal sync server that doesn't include Anki's GUI dependencies. Python and Rust implementations are available.

From a Packaged Build

This uses the sync server built into the desktop version of Anki as of version 2.1.57+.

On Windows in a cmd.exe session:

set SYNC_USER1=user:pass
"\Program Files\anki\anki-console" --syncserver

Or MacOS, in Terminal.app:

SYNC_USER1=user:pass /Applications/Anki.app/Contents/MacOS/launcher --syncserver

Replace 'launcher' with 'anki' for old packaged builds prior to 25.07.

Or Linux:

SYNC_USER1=user:pass anki --syncserver

With Pip

To avoid downloading desktop Anki's GUI dependencies, you can run a standalone Anki sync server using a Python package downloaded from PyPI instead. Make sure you have Python 3.9+ installed.

python3 -m venv ~/syncserver
~/syncserver/bin/pip install anki
SYNC_USER1=user:pass ~/syncserver/bin/python -m anki.syncserver

With Cargo

From Anki 2.1.66+, you can alternatively build a Rust implementation of the standalone sync server using the below command. Make sure you have Rustup installed.

cargo install --locked --git https://github.com/ankitects/anki.git --tag 25.02.5 anki-sync-server

Replace 25.02.5 with whatever the latest Anki version is.

Protobuf (protoc) will need to be installed.

After building, you can run it with:

SYNC_USER1=user:pass anki-sync-server

From a source checkout

If you've cloned the Anki repo from GitHub, you can install from there:

./ninja extract:protoc ftl_repo
cargo install --path rslib/sync

With Docker

You can find a user-contributed Dockerfile and some instructions here.

Multiple Users

SYNC_USER1 declares the first user and password, and must be set. You can optionally declare SYNC_USER2, SYNC_USER3 and so on, if you wish to set up multiple accounts.

Hashed Passwords

Advanced users may wish to use hashed passwords instead of plain text passwords. If you wish to do this, you'll need to use a separate tool (such as this one) to generate a password hash. You can then tell the server to expect hashed passwords by setting the env var PASSWORDS_HASHED to 1 (or any other value).

When hashed passwords are used, SYNC_USER variables are expected to be in username:password_hash format, where password_hash is a hash of the password in the PHC Format.

Storage Location

The server needs to store a copy of your collection and media in a folder. By default it is ~/.syncserver; you can change this by defining a SYNC_BASE environment variable.

• This must not be the same location as your normal Anki data folder, as the server and client must store separate copies.
• You must sync your data to the server, not manually copy files into the server folder.

Public Access

The server listens on an unencrypted HTTP connection, so it's not a good idea to expose it directly to the internet. You'll want to either restrict usage to your local network, or place some form of encryption in front of the server, such as a VPN (Tailscale is apparently easy), or a HTTPS reverse proxy.

You can define SYNC_HOST and SYNC_PORT to change the host and port that the server binds to.

Client Setup

You'll need to determine your computer's network IP address, and then point each of your Anki clients to the address, e.g something like http://192.168.1.200:8080/. The URL can be configured in the preferences.

If you're using AnkiMobile and are unable to connect to a server on your local network, please go into the iOS settings, locate Anki near the bottom, and toggle "Allow Anki to access local network" off and then on again.

Older desktop clients required you to define SYNC_ENDPOINT and SYNC_ENDPOINT_MEDIA. If using an older client, you'd put it as e.g. http://192.168.1.200:8080/sync/ and http://192.168.1.200:8080/msync/ respectively. AnkiDroid clients before 2.16 require separate configuration for the two endpoints.

Reverse Proxies

If using a reverse proxy to provide HTTPS access (e.g. nginx), and binding to a subpath (e.g. http://example.com/custom/ -> http://localhost:8080/), you must make sure to include a trailing slash when configuring Anki. If you put http://example.com/custom instead, it will not work.

If you are using a caddy, increase the http read buffer size to avoid connection problems when downloading media files.

reverse_proxy http://127.0.0.1:8080 {
	transport http {
		read_buffer 512k
	}
}

On iOS, TLS 1.3 is not supported, so your reverse proxy will need to have TLS 1.2 enabled, or you'll get an "error code -9836".

Large Requests

The standard AnkiWeb limit on uploads is applied by default. You can optionally set MAX_SYNC_PAYLOAD_MEGS to something greater than 100 if you wish to increase the limit. Bear in mind that if you're using a reverse proxy, you may need to adjust the limit there as well.

Contributing Changes

Because this server is bundled with Anki, simplicity is a design goal - it is targeted at individual/family use, and PRs that add things like a REST API or external databases are unlikely to be accepted at this time. If in doubt, please reach out before starting work on a PR.

If you're looking for an existing API solution, the AnkiConnect add-on may meet your needs.