Build messaging apps for Android Auto

The Communications category is coming soon

The Messaging category is being expanded to include support for new capabilities, including message history and calling experiences

Nominate yourself to be an early access partner →

Staying connected through messages is important to many drivers. Chat apps can let users know if a child needs to be picked up or if a dinner location has been changed. The Android framework lets messaging apps extend their services into the driving experience using a standard user interface that lets drivers keep their eyes on the road.

Apps that support messaging can extend their messaging notifications to let Android Auto consume them when Auto is running. These notifications are displayed in Auto and let users read and respond to messages in a consistent, low-distraction interface. And when you use the MessagingStyle API, you get optimized message notifications for all Android devices, including Android Auto. The optimizations include a UI that's specialized for message notifications, improved animations, and support for inline images.

This guide shows you how to extend an app that displays messages to the user and receives the user's replies, such as a chat app, to hand message display and reply receipt off to an Auto device. For related design guidance, see Messaging apps on the Design for Driving site.

Get started

To provide messaging service for Auto devices, your app must declare its support for Android Auto in the manifest and be able to do the following:

• Build and send NotificationCompat.MessagingStyle objects that contain reply and mark-as-read Action objects.
• Handle replying and marking a conversation as read with a Service.

Concepts and objects

Before you start designing your app, it's helpful to understand how Android Auto handles messaging.

An individual chunk of communication is called a message and is represented by the class MessagingStyle.Message. A message contains a sender, the message content, and the time the message was sent.

Communication between users is called a conversation and is represented by a MessagingStyle object. A conversation, or MessagingStyle, contains a title, the messages, and whether the conversation is among a group of users.

To notify users of updates to a conversation, such as a new message, apps post a Notification to the Android system. This Notification uses the MessagingStyle object to display messaging-specific UI in the notification shade. The Android platform also passes this Notification to Android Auto, and the MessagingStyle is extracted and used to post a notification through the car's display.

Android Auto also requires apps to add Action objects to a Notification to let the user quickly reply to a message or mark it as read directly from the notification shade.

In summary, a single conversation is represented by a Notification object that is styled with a MessagingStyle object. The MessagingStyle contains all the messages within that conversation in one or more MessagingStyle.Message objects. And, to be Android Auto compliant, an app must attach reply and mark-as-read Action objects to the Notification.

Messaging flow

This section describes a typical messaging flow between your app and Android Auto.

1. Your app receives a message.
2. Your app generates a MessagingStyle notification with reply and mark-as-read Action objects.
3. Android Auto receives the “new notification” event from the Android system and finds the MessagingStyle, reply Action, and mark-as-read Action.
4. Android Auto generates and displays a notification in the car.
5. If the user taps the notification on the car's display, Android Auto triggers the mark-as-read Action.
   • In the background, your app must handle this mark-as-read event.
6. If the user responds to the notification using voice, Android Auto puts a transcription of the user's response into the reply Action and then triggers it.
   • In the background, your app must handle this reply event.

Preliminary assumptions

This page does not guide you in creating an entire messaging app. The following code sample includes some of the things your app needs before you start to support messaging with Android Auto:

data class YourAppConversation(
        val id: Int,
        val title: String,
        val recipients: MutableList<YourAppUser>,
        val icon: Bitmap) {
    companion object {
        /** Fetches [YourAppConversation] by its [id]. */
        fun getById(id: Int): YourAppConversation = // ...
    }

    /** Replies to this conversation with the given [message]. */
    fun reply(message: String) {}

    /** Marks this conversation as read. */
    fun markAsRead() {}

    /** Retrieves all unread messages from this conversation. */
    fun getUnreadMessages(): List<YourAppMessage> { return /* ... */ }
}
data class YourAppUser(val id: Int, val name: String, val icon: Uri)
data class YourAppMessage(
    val id: Int,
    val sender: YourAppUser,
    val body: String,
    val timeReceived: Long)

Declare Android Auto support

When Android Auto receives a notification from a messaging app, it checks that the app has declared support for Android Auto. To enable this support, include the following entry in your app's manifest:

<application>
    ...
    <meta-data
        android:name="com.google.android.gms.car.application"
        android:resource="@xml/automotive_app_desc"/>
    ...
</application>

This manifest entry refers to another XML file you need to create with the following path: YourAppProject/app/src/main/res/xml/automotive_app_desc.xml. In automotive_app_desc.xml declare the Android Auto capabilities your app supports. For example, to declare support for notifications, include the following:

<automotiveApp>
    <uses name="notification" />
</automotiveApp>

If your app can be set as the default SMS handler, make sure to include the following <uses> element. If you do not, a default handler built-in to Android Auto will be used to handle incoming SMS/MMS messages when your app is set as the default SMS handler, which can lead to duplicate notifications.

<automotiveApp>
    ...
    <uses name="sms" />
</automotiveApp>

Import the AndroidX core library

Building notifications for use with Auto devices requires the AndroidX core library. Import the library into your project as follows:

1. In the top-level build.gradle file, include a dependency on Google's Maven repository, as shown in the following example:

allprojects {
    repositories {
        google()
    }
}

allprojects {
    repositories {
        google()
    }
}

1. In your app module's build.gradle file, include the AndroidX Core library dependency, as shown in the following example:

dependencies {
    // If your app is written in Java
    implementation 'androidx.core:core:1.15.0'

    // If your app is written in Kotlin
    implementation 'androidx.core:core-ktx:1.15.0'
}

dependencies {
    // If your app is written in Java
    implementation("androidx.core:core:1.15.0")

    // If your app is written in Kotlin
    implementation("androidx.core:core-ktx:1.15.0")
}

Handle user actions

Your messaging app needs a way to handle updating a conversation through an Action. For Android Auto, there are two types of Action objects your app needs to handle: reply and mark-as-read. We recommend handling them using an IntentService, which provides the flexibility to handle potentially expensive calls in the background, freeing your app's main thread.

Define intent actions

Intent actions are simple strings that identify what the Intent is for. Because a single service can handle multiple types of intents, it's easier to define multiple action strings instead of defining multiple IntentService components.

This guide's example messaging app has the two required types of actions: reply and mark-as-read, as shown in the following code sample.

private const val ACTION_REPLY = "com.example.REPLY"
private const val ACTION_MARK_AS_READ = "com.example.MARK_AS_READ"

Create the service

To create a service that handles these Action objects, you need the conversation ID, which is an arbitrary data structure defined by your app that identifies the conversation. You also need a remote input key, which is discussed in detail later in this section. The following code sample creates a service to handle the required actions:

private const val EXTRA_CONVERSATION_ID_KEY = "conversation_id"
private const val REMOTE_INPUT_RESULT_KEY = "reply_input"

/**
 * An [IntentService] that handles reply and mark-as-read actions for
 * [YourAppConversation]s.
 */
class MessagingService : IntentService("MessagingService") {
    override fun onHandleIntent(intent: Intent?) {
        // Fetches internal data.
        val conversationId = intent!!.getIntExtra(EXTRA_CONVERSATION_ID_KEY, -1)

        // Searches the database for that conversation.
        val conversation = YourAppConversation.getById(conversationId)

        // Handles the action that was requested in the intent. The TODOs
        // are addressed in a later section.
        when (intent.action) {
            ACTION_REPLY -> TODO()
            ACTION_MARK_AS_READ -> TODO()
        }
    }
}

To associate this service with your app, you also need to register the service in your app's manifest, as shown in the following example:

<application>
    <service android:name="com.example.MessagingService" />
    ...
</application>

Generate and handle intents

There's no way for other apps, including Android Auto, to obtain the Intent that triggers the MessagingService, because Intents are passed to other apps through a PendingIntent. Because of this limitation, you need to create a RemoteInput object to let other apps provide the reply text back to your app, as shown in the following example:

/**
 * Creates a [RemoteInput] that lets remote apps provide a response string
 * to the underlying [Intent] within a [PendingIntent].
 */
fun createReplyRemoteInput(context: Context): RemoteInput {
    // RemoteInput.Builder accepts a single parameter: the key to use to store
    // the response in.
    return RemoteInput.Builder(REMOTE_INPUT_RESULT_KEY).build()
    // Note that the RemoteInput has no knowledge of the conversation. This is
    // because the data for the RemoteInput is bound to the reply Intent using
    // static methods in the RemoteInput class.
}

/** Creates an [Intent] that handles replying to the given [appConversation]. */
fun createReplyIntent(
        context: Context, appConversation: YourAppConversation): Intent {
    // Creates the intent backed by the MessagingService.
    val intent = Intent(context, MessagingService::class.java)

    // Lets the MessagingService know this is a reply request.
    intent.action = ACTION_REPLY

    // Provides the ID of the conversation that the reply applies to.
    intent.putExtra(EXTRA_CONVERSATION_ID_KEY, appConversation.id)

    return intent
}

In the ACTION_REPLY switch clause within the MessagingService, extract the information that goes into the reply Intent, as shown in the following example:

ACTION_REPLY -> {
    // Extracts reply response from the intent using the same key that the
    // RemoteInput uses.
    val results: Bundle = RemoteInput.getResultsFromIntent(intent)
    val message = results.getString(REMOTE_INPUT_RESULT_KEY)

    // This conversation object comes from the MessagingService.
    conversation.reply(message)
}

You handle the mark-as-read Intent in a similar way. However, it doesn't require a RemoteInput, as shown in the following example:

/** Creates an [Intent] that handles marking the [appConversation] as read. */
fun createMarkAsReadIntent(
        context: Context, appConversation: YourAppConversation): Intent {
    val intent = Intent(context, MessagingService::class.java)
    intent.action = ACTION_MARK_AS_READ
    intent.putExtra(EXTRA_CONVERSATION_ID_KEY, appConversation.id)
    return intent
}

The ACTION_MARK_AS_READ switch clause within the MessagingService requires no further logic, as shown in the following example:

// Marking as read has no other logic.
ACTION_MARK_AS_READ -> conversation.markAsRead()

Notify users of messages

Once conversation action handling is complete, the next step is to generate Android Auto compliant notifications.

Create actions

Action objects can be passed to other apps using a Notification to trigger methods in the original app. This is how Android Auto can mark a conversation as read or reply to it.

To create an Action, start with an Intent. The following example shows how to create a "reply" Intent:

fun createReplyAction(
        context: Context, appConversation: YourAppConversation): Action {
    val replyIntent: Intent = createReplyIntent(context, appConversation)
    // ...

Then, wrap this Intent in a PendingIntent, which prepares it for external app usage. A PendingIntent locks down all access to the wrapped Intent by only exposing a select set of methods that let the receiving app fire the Intent or get the originating app's package name. The external app can't ever access the underlying Intent or the data within it.

    // ...
    val replyPendingIntent = PendingIntent.getService(
        context,
        createReplyId(appConversation), // Method explained later.
        replyIntent,
        PendingIntent.FLAG_UPDATE_CURRENT or PendingIntent.FLAG_MUTABLE)
    // ...

Before you set up the reply Action, be aware that Android Auto has three requirements for the reply Action:

• The semantic action must be set to Action.SEMANTIC_ACTION_REPLY.
• The Action must indicate that it will not show any user interface when fired.
• The Action must contain a single RemoteInput.

The following code sample sets up a reply Action that addresses the requirements listed above:

    // ...
    val replyAction = Action.Builder(R.drawable.reply, "Reply", replyPendingIntent)
        // Provides context to what firing the Action does.
        .setSemanticAction(Action.SEMANTIC_ACTION_REPLY)

        // The action doesn't show any UI, as required by Android Auto.
        .setShowsUserInterface(false)

        // Don't forget the reply RemoteInput. Android Auto will use this to
        // make a system call that will add the response string into
        // the reply intent so it can be extracted by the messaging app.
        .addRemoteInput(createReplyRemoteInput(context))
        .build()

    return replyAction
}

Handling the mark-as-read action is similar, except there's no RemoteInput. Android Auto therefore has two requirements for the mark-as-read Action:

• The semantic action is set to Action.SEMANTIC_ACTION_MARK_AS_READ.
• The action indicates that it will not show any user interface when fired.

The following code sample sets up a mark-as-read Action that addresses these requirements:

fun createMarkAsReadAction(
        context: Context, appConversation: YourAppConversation): Action {
    val markAsReadIntent = createMarkAsReadIntent(context, appConversation)
    val markAsReadPendingIntent = PendingIntent.getService(
            context,
            createMarkAsReadId(appConversation), // Method explained below.
            markAsReadIntent,
            PendingIntent.FLAG_UPDATE_CURRENT  or PendingIntent.FLAG_IMMUTABLE)
    val markAsReadAction = Action.Builder(
            R.drawable.mark_as_read, "Mark as Read", markAsReadPendingIntent)
        .setSemanticAction(Action.SEMANTIC_ACTION_MARK_AS_READ)
        .setShowsUserInterface(false)
        .build()
    return markAsReadAction
}

When generating the pending intents, two methods are used: createReplyId() and createMarkAsReadId(). These methods serve as the request codes for each PendingIntent, which are used by Android to control existing pending intents. The create() methods must return unique IDs for each conversation, but repeated calls for the same conversation must return the unique ID already generated.

Consider an example with two conversations, A and B: Conversation A's reply ID is 100, and its mark-as-read ID is 101. Conversation B's reply ID is 102, and its mark-as-read ID is 103. If conversation A is updated, the reply and mark-as-read IDs are still 100 and 101. For more information, see PendingIntent.FLAG_UPDATE_CURRENT.

Create a MessagingStyle

MessagingStyle is the carrier of the messaging information and is what Android Auto uses to read aloud each message in a conversation.

First, the user of the device must be specified in the form of a Person object, as shown in the following example:

fun createMessagingStyle(
        context: Context, appConversation: YourAppConversation): MessagingStyle {
    // Method defined by the messaging app.
    val appDeviceUser: YourAppUser = getAppDeviceUser()

    val devicePerson = Person.Builder()
        // The display name (also the name that's read aloud in Android auto).
        .setName(appDeviceUser.name)

        // The icon to show in the notification shade in the system UI (outside
        // of Android Auto).
        .setIcon(appDeviceUser.icon)

        // A unique key in case there are multiple people in this conversation with
        // the same name.
        .setKey(appDeviceUser.id)
        .build()
    // ...

You can then construct the MessagingStyle object and provide some details about the conversation.

    // ...
    val messagingStyle = MessagingStyle(devicePerson)

    // Sets the conversation title. If the app's target version is lower
    // than P, this will automatically mark the conversation as a group (to
    // maintain backward compatibility). Use `setGroupConversation` after
    // setting the conversation title to explicitly override this behavior. See
    // the documentation for more information.
    messagingStyle.setConversationTitle(appConversation.title)

    // Group conversation means there is more than 1 recipient, so set it as such.
    messagingStyle.setGroupConversation(appConversation.recipients.size > 1)
    // ...

Finally, add the unread messages.

    // ...
    for (appMessage in appConversation.getUnreadMessages()) {
        // The sender is also represented using a Person object.
        val senderPerson = Person.Builder()
            .setName(appMessage.sender.name)
            .setIcon(appMessage.sender.icon)
            .setKey(appMessage.sender.id)
            .build()

        // Adds the message. More complex messages, like images,
        // can be created and added by instantiating the MessagingStyle.Message
        // class directly. See documentation for details.
        messagingStyle.addMessage(
                appMessage.body, appMessage.timeReceived, senderPerson)
    }

    return messagingStyle
}

Package and push the notification

After generating the Action and MessagingStyle objects, you can construct and post the Notification.

fun notify(context: Context, appConversation: YourAppConversation) {
    // Creates the actions and MessagingStyle.
    val replyAction = createReplyAction(context, appConversation)
    val markAsReadAction = createMarkAsReadAction(context, appConversation)
    val messagingStyle = createMessagingStyle(context, appConversation)

    // Creates the notification.
    val notification = NotificationCompat.Builder(context, channel)
        // A required field for the Android UI.
        .setSmallIcon(R.drawable.notification_icon)

        // Shows in Android Auto as the conversation image.
        .setLargeIcon(appConversation.icon)

        // Adds MessagingStyle.
        .setStyle(messagingStyle)

        // Adds reply action.
        .addAction(replyAction)

        // Makes the mark-as-read action invisible, so it doesn't appear
        // in the Android UI but the app satisfies Android Auto's
        // mark-as-read Action requirement. Both required actions can be made
        // visible or invisible; it is a stylistic choice.
        .addInvisibleAction(markAsReadAction)

        .build()

    // Posts the notification for the user to see.
    val notificationManagerCompat = NotificationManagerCompat.from(context)
    notificationManagerCompat.notify(appConversation.id, notification)
}

Additional resources

• Google Design for Driving: Messaging apps
• Notifications overview

Report an Android Auto Messaging issue

If you run into an issue while developing your messaging app for Android Auto, you can report it using the Google Issue Tracker. Be sure to fill out all the requested information in the issue template.

Create a new issue

Before filing a new issue, check whether it is already reported in the issues list. You can subscribe and vote for issues by clicking the star for an issue in the tracker. For more information, see Subscribing to an Issue.