Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?


Failed to load latest commit information.
Latest commit message
Commit time
March 1, 2020 14:21
January 14, 2023 20:35
March 15, 2020 12:23
January 14, 2023 20:35
January 14, 2023 20:35

Linux Build Status Mac Build Status Windows Build Status

Maven Central

Multiplatform Settings

This is a Kotlin library for Multiplatform apps, so that common code can persist key-value data.

Table of contents

Adding to your project

Multiplatform Settings is currently published to Maven Central, so add that to repositories.

repositories {
    // ...

Then, simply add the dependency to your common source-set dependencies

commonMain {
    dependencies {
        // ...

See also the sample project, which uses this structure.


The Settings interface has implementations on the Android, iOS, macOS, watchOS, tvOS, JS, JVM, and Windows platforms.

Creating a Settings instance

When writing multiplatform code, you might need to interoperate with platform-specific code which needs to share the same data-source. To facilitate this, all Settings implementations wrap a delegate object which you could also use in your platform code.

Since that delegate is a constructor argument, it should be possible to connect it via any dependency-injection strategy you might already be using. If your project doesn't have such a system already in place, one strategy is to use expect declarations, for example

expect val settings: Settings

// or
expect fun createSettings(): Settings

Then the actual implementations can pass the platform-specific delegates. See Platform constructors below for more details on these delegates.

Some platform implementations also include Factory classes. These make it easier to manage multiple named Settings objects from common code, or to automate some platform-specific configuration so that delegates don't need to be created manually. The factory still needs to be injected from platform code, but then from common you can call

val settings1: Settings = factory.create("my_first_settings")
val settings2: Settings = factory.create("my_other_settings")

See Factories below for more details.

However, if all of your key-value logic exists in a single instance in common code, these ways of instantiation Settings can be inconvenient. To make pure-common usage easier, Multiplatform Settings now includes a separate module which provides a Settings() factory function, so that you can create a Settings instance like

val settings: Settings = Settings()

See No-arg module below for more details.

Platform constructors

The Android implementation is SharedPreferencesSettings, which wraps SharedPreferences.

val delegate: SharedPreferences // ...
val settings: Settings = SharedPreferencesSettings(delegate)

On iOS, macOS, tvOS, or watchOS, NSUserDefaultsSettings wraps NSUserDefaults.

val delegate: NSUserDefaults // ...
val settings: Settings = NSUserDefaultsSettings(delegate)

You can also use KeychainSettings which writes to the Keychain. Construct it by passing a String which will be interpreted as a service name.

val serviceName: String // ...
val settings: Settings = KeychainSettings(serviceName)

Two JVM implementations exist. PreferencesSettings wraps Preferences and PropertiesSettings wraps Properties.

val delegate: Preferences // ...
val settings: Settings = PreferencesSettings(delegate)

val delegate: Properties // ...
val settings: Settings = PropertiesSettings(delegate)

On JS, JsSettings wraps Storage.

val delegate: Storage // ...
val settings: Settings = JsSettings(delegate)

val settings: Settings = JsSettings() // use localStorage by default

There is a Windows implementation RegistrySettings which wraps the Windows registry.

val rootKey: String = "SOFTWARE\\..." // Will be interpreted as subkey of HKEY_CURRENT_USER
val settings: Settings = RegistrySettings(rootKey)


For some platforms, a Factory class also exists, so that multiple named Settings instances can coexist with the names being controlled from common code.

On Android, this factory needs a Context parameter

val context: Context // ...
val factory: Settings.Factory = SharedPreferencesSettings.Factory(context)

On most other platforms, the factory can be instantiated without passing any parameter

val factory: Settings.Factory = NSUserDefaultsSettings.Factory()

If you have a Factory reference from your common code, then you can use it to create multiple Settings with different names.

val settings1: Settings = factory.create("my_first_settings")
val settings2: Settings = factory.create("my_other_settings")

If the default Factorys don't do what you need, you can also implement your own.

No-arg module

To create a Settings instance from common without needing to pass platform-specific dependencies, add the multiplatform-settings-no-arg gradle dependency. This exports multiplatform-settings as an API dependency, so you can use it as a replacement for that default dependency.


Then from common code, you can write

val settings: Settings = Settings()

This is implemented via an extension function operator fun Settings.Companion.invoke() to provide constructor-like syntax even though Settings has no constructor.

On Android, this delegates to the equivalent of PreferenceManager.getDefaultSharedPreferences() internally. It makes use of a content-provider to get a context reference without needing to pass one manually. On Apple platforms, it uses NSUserDefaults.standardUserDefaults. On JS, it uses localStorage. On JVM, it uses the Preferences implementation with Preferences.userRoot() as a delegate. On Windows, it reads the name of the executable being built and writes to a subkey of HKEY_CURRENT_USER\SOFTWARE using that name.

Note that while the main multiplatform-settings module publishes common code to all available Kotlin platforms, the multiplatform-settings-no-arg module only publishes to platforms which have concrete implementations.

Note also that the no-arg module is there to make getting started easier with less configuration, but there are plenty of things it doesn't provide, such as the ability to use an encrypted implementation on platforms that support it, or the ability to substitute a test implementation. If you need a non-default setup you likely are better off not using multiplatform-settings-no-arg.

Settings API

Once the Settings instance is created, you can store values by calling the various putXXX() methods, or their operator shortcuts

settings.putInt("key", 3)
settings["key"] = 3

You can retrieve stored values via the getXXX() methods or their operator shortcuts. If a key is not present, then the supplied default will be returned instead.

val a: Int = settings.getInt("key")
val b: Int = settings.getInt("key", defaultValue = -1)
val c: Int = settings["key", -1]

Nullable methods are also available to avoid the need to use a default value. Instead, null will be returned if a key is not present.

val a: Int? = settings.getIntOrNull("key")
val b: Int? = settings["key"]

The getXXX() and putXXX() operation for a given key can be wrapped using a property delegate. This has the advantage of ensuring that the key is always accessed with a consistent type.

val a: Int by"key")
val b: Int by"key", defaultValue = -1)

Nullable delegates exists so that absence of a key can be indicated by null instead of a default value

val a: Int? by settings.nullableInt("key")

The key parameter can be omitted for delegates, and the property name will be reflectively used instead.

val a: Int by // internally, key is "a"

Existence of a key can be queried

val a: Boolean = settings.hasKey("key")
val b: Boolean = "key" in settings

Values can also be removed by key

settings -= "key"
settings["key"] = null

Finally, all values in a Settings instance can be removed


The set of keys and amount of entries can be retrieved

val keys: Set<String> = settings.keys
val size: Int = settings.size

Note that for the NSUserDefaultsSettings implementation, some entries are unremovable and therefore may still be present after a clear() call. Thus, size is not generally guaranteed to be zero after a clear().


Update listeners are available for some implementations. These are marked with the ObservableSettings interface, which includes an addListener() method.

val observableSettings: ObservableSettings // ...
val settingsListener: SettingsListener = observableSettings.addIntListener(key) { value: Int -> /* ... */ }
val settingsListener: SettingsListener = observableSettings.addNullableIntListener(key) { value: Int? -> /* ... */ }

The SettingsListener returned from the call should be used to signal when you're done listening:


If you don't hold a strong reference to the SettingsListener, it's possible in some implementations that it will be garbage-collected and stop sending updates.


A testing dependency is available to aid in testing code that interacts with this library.


This includes a MapSettings implementation of the Settings interface, which is backed by an in-memory MutableMap on all platforms.

Other platforms

The Settings interface is published to all available platforms. Developers who desire implementations outside of the defaults provided are free to add their own implementations, and are welcome to make pull requests if the implementation might be generally useful to others. Note that implementations which require external dependencies should be places in a separate gradle module in order to keep the core multiplatform-settings module dependency-free.

Experimental API

Certain APIs are marked with @ExperimentalSettingsApi or @ExperimentalSettingsImplementation to highlight areas that may have the potential to break in the future and should not be considered stable to depend on.

Experimental Implementations

Apple Keychain

The KeychainSettings implementation on Apple platforms and the RegistrySettings implementation on Windows are considered experimental. Feel free to reach out if they're working well for you, or if you encounter any issues with them, to help remove that experimental status.

Serialization module

A kotlinx-serialization integration exists so it's easier to save non-primitive data


This essentially uses the Settings store as a serialization format. Thus for a serializable class

class SomeClass(val someProperty: String, anotherProperty: Int)

an instance can be stored or retrieved

val someClass: SomeClass
val settings: Settings

// Store values for the properties of someClass in settings
settings.encodeValue(SomeClass.serializer(), "key", someClass)

// Create a new instance of SomeClass based on the data in settings
val newInstance: SomeClass = settings.decodeValue(SomeClass.serializer(), "someClass", defaultValue)
val nullableNewInstance: SomeClass = settings.decodeValueOrNull(SomeClass.serializer(), "someClass")

There's also a delegate API, similar to that for primitives

val someClass: SomeClass by settings.serializedValue(SomeClass.serializer(), "someClass", defaultValue)
val nullableSomeClass: SomeClass? by settings.nullableSerializedValue(SomeClass.serializer(), "someClass")

Usage requires accepting both the @ExperimentalSettingsApi and @ExperimentalSerializationApi annotations.

Coroutine APIs

A separate multiplatform-settings-coroutines dependency includes various coroutine APIs.


This adds flow extensions for all types which use the listener APIs internally.

val observableSettings: ObservableSettings // Only works with ObservableSettings
val flow: Flow<Int> by observableSettings.intFlow("key", defaultValue)
val nullableFlow: Flow<Int?> by observableSettings.intOrNullFlow("key")

In addition, there are two new Settings-like interfaces: SuspendSettings, which looks similar to Settings but all functions are marked suspend, and FlowSettings which extends SuspendSettings to also include Flow-based getters similar to the extensions mentioned above.

val suspendSettings: SuspendSettings // ...
val a: Int = suspendSettings.getInt("key") // This call will suspend

val flowSettings: FlowSettings // ...
val flow: Flow<Int> = flowSettings.getIntFlow("key")

There are APIs provided to convert between these different interfaces so that you can select one to use primarily from common.

val settings: Settings // ...
val suspendSettings: SuspendSettings = settings.toSuspendSettings()

val observableSettings: ObservableSettings // ...
val flowSettings: FlowSettings = observableSettings.toFlowSettings()

// Wrap suspend calls in runBlocking
val blockingSettings: Settings = suspendSettings.toBlockingSettings()
val blockingSettings: ObservableSettings = flowSettings.toBlockingObservableSettings()


An implementation of FlowSettings on the Android exists in the multiplatform-settings-datastore dependency, based on Jetpack DataStore


This provides a DataStoreSettings class

val dataStore: DataStore // = ...
val settings: FlowSettings = DataStoreSettings(dataStore)

You can use this in shared code by converting other ObservableSettings instances to FlowSettings. For example:

// Common
expect val settings: FlowSettings

// Android
actual val settings: FlowSettings = DataStoreSettings(/*...*/)

// iOS
actual val settings: FlowSettings = NSUserDefaultsSettings(/*...*/).toFlowSettings()

Or, if you also include platforms without listener support, you can use SuspendSettings instead.

// Common
expect val settings: SuspendSettings

// Android
actual val settings: SuspendSettings = DataStoreSettings(/*...*/)

// iOS
actual val settings: SuspendSettings = NSUserDefaultsSettings(/*...*/).toSuspendSettings()

// JS
actual val settings: SuspendSettings = JsSettings().toSuspendSettings()


The project includes multiple CI jobs configured using Github Actions. On PRs or updates to the main branch, the build will run the scripts in build-linux.yml, build-macos.yml, build-windows.yml, and validate-gradle-wrapper.yml. These builds the library and runs unit tests for all platforms across Linux, Mac, and Windows hosts. In addition, the library build artifacts are deployed to the local maven repository and the sample project is built for the platforms on which it is implemented. This ensures that the sample remains in sync with updates to the library.

An addition build script is defined in deploy.yml, which runs on a manual trigger. This builds the library for all platforms and uploads artifacts to staging on Maven Central. Uploaded artifacts must still be published manually


Copyright 2018-2023 Russell Wolf

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
See the License for the specific language governing permissions and
limitations under the License.

Jetbrains Logo

Made with JetBrains tools