Getting Started with the sbt launcher 

The sbt launcher provides two parts:

  1. An interface for launched applications to interact with the launcher code
  2. A minimal sbt-launch.jar that can launch applications by resolving them through ivy.

The sbt launcher component is a self-contained jar that boots a Scala application or server without Scala or the application already existing on the system. The only prerequisites are the launcher jar itself, an optional configuration file, and a Java runtime version 1.6 or greater.


A user downloads the launcher jar and creates a script to run it. In this documentation, the script will be assumed to be called launch. For Unix, the script would look like: java -jar sbt-launcher.jar "$@"

The user can now launch servers and applications which provide sbt launcher configuration.

Alternatively, you can repackage the launcher with a launcher configuration file. For example, sbt/sbt pulls in the raw JAR and injects the appropriate files for sbt.


To launch an application, the user then downloads the configuration file for the application (call it and creates a script to launch it (call it myapp):

launch "$@"

The user can then launch the application using myapp arg1 arg2 ...

More on launcher configuration can be found at Launcher Configuration


The sbt launcher can be used to launch and discover running servers on the system. The launcher can be used to launch servers similarly to applications. However, if desired, the launcher can also be used to ensure that only one instance of a server is running at time. This is done by having clients always use the launcher as a service locator.

To discover where a server is running (or launch it if it is not running), the user downloads the configuration file for the server (call it my.server.configuration) and creates a script to discover the server (call it find-myserver):

launch --locate

This command will print out one string, the URI at which to reach the server, e.g. sbt:// Clients should use the IP/port to connect to to the server and initiate their connection.

When using the locate feature, the sbt launcher makes the following restrictions to servers:

Resolving Applications/Servers 

Like the launcher used to distribute sbt, the downloaded launcher jar will retrieve Scala and the application according to the provided configuration file. The versions may be fixed or read from a different configuration file (the location of which is also configurable). The location to which the Scala and application jars are downloaded is configurable as well. The repositories searched are configurable. Optional initialization of a properties file on launch is configurable.

Once the launcher has downloaded the necessary jars, it loads the application/server and calls its entry point. The application is passed information about how it was called: command line arguments, current working directory, Scala version, and application ID (organization, name, version). In addition, the application can ask the launcher to perform operations such as obtaining the Scala jars and a ClassLoader for any version of Scala retrievable from the repositories specified in the configuration file. It can request that other applications be downloaded and run. When the application completes, it can tell the launcher to exit with a specific exit code or to reload the application with a different version of Scala, a different version of the application, or different arguments.

There are some other options for setup, such as putting the configuration file inside the launcher jar and distributing that as a single download. The rest of this documentation describes the details of configuring, writing, distributing, and running the application.

Creating a Launched Application 

This section shows how to make an application that is launched by this launcher. First, declare a dependency on the launcher-interface. Do not declare a dependency on the launcher itself. The launcher interface consists strictly of Java interfaces in order to avoid binary incompatibility between the version of Scala used to compile the launcher and the version used to compile your application. The launcher interface class will be provided by the launcher, so it is only a compile-time dependency. If you are building with sbt, your dependency definition would be:

libraryDependencies += "org.scala-sbt" % "launcher-interface" % "1.0.0" % "provided"

resolvers += sbtResolver.value

Make the entry point to your class implement xsbti.AppMain. An example that uses some of the information:

package com.acme.launcherapp

class Main extends xsbti.AppMain
    def run(configuration: xsbti.AppConfiguration) =
        // get the version of Scala used to launch the application
        val scalaVersion = configuration.provider.scalaProvider.version

        // Print a message and the arguments to the application
        println("Hello world!  Running Scala " + scalaVersion)

        // demonstrate the ability to reboot the application into different versions of Scala
        // and how to return the code to exit with
        scalaVersion match
            case "2.10.6" =>
                new xsbti.Reboot {
                    def arguments = configuration.arguments
                    def baseDirectory = configuration.baseDirectory
                    def scalaVersion = "2.11.8"
                    def app =
            case "2.11.8" => new Exit(1)
            case _ => new Exit(0)
    class Exit(val code: Int) extends xsbti.Exit

Next, define a configuration file for the launcher. For the above class, it might look like:

  version: 2.11.8
  org: com.acme
  name: launcherapp
  version: 0.0.1
  class: com.acme.launcherapp.Main
  cross-versioned: true
 directory: ${user.home}/.myapp/boot

Then, publishLocal or +publishLocal the application in sbt’s shell to make it available. For more information, see Launcher Configuration.

Running an Application 

As mentioned above, there are a few options to actually run the application. The first involves providing a modified jar for download. The second two require providing a configuration file for download.


Let’s review what’s happening when the launcher starts your application.

On startup, the launcher searches for its configuration and then parses it. Once the final configuration is resolved, the launcher proceeds to obtain the necessary jars to launch the application. The property is used as a base directory to retrieve jars to. Locking is done on the directory, so it can be shared system-wide. The launcher retrieves the requested version of Scala to


If this directory already exists, the launcher takes a shortcut for startup performance and assumes that the jars have already been downloaded. If the directory does not exist, the launcher uses Apache Ivy to resolve and retrieve the jars. A similar process occurs for the application itself. It and its dependencies are retrieved to


Once all required code is downloaded, the class loaders are set up. The launcher creates a class loader for the requested version of Scala. It then creates a child class loader containing the jars for the requested app.components and with the paths specified in app.resources. An application that does not use components will have all of its jars in this class loader.

The main class for the application is then instantiated. It must be a public class with a public no-argument constructor and must conform to xsbti.AppMain. The run method is invoked and execution passes to the application. The argument to the ‘run’ method provides configuration information and a callback to obtain a class loader for any version of Scala that can be obtained from a repository in [repositories]. The return value of the run method determines what is done after the application executes. It can specify that the launcher should restart the application or that it should exit with the provided exit code.


sbt Reference Manual
      1. Getting Started with the sbt launcher