JetBrains now recommends Kotlin Toolchain 0.13 for most new multiplatform projects

The most useful fix in 0.13 is a small one. Before this release, the Toolchain provisioned the Android SDK when a developer ran `kotlin build` or `kotlin run`, but not when the project was opened in an IDE ^[12]. That meant a fresh checkout behaved differently depending on which entry point the developer used first. Open the project before building and missing SDKs could cause setup errors ^[12]. JetBrains aligned the CLI and IDE behaviour, and the Toolchain now provisions missing Android SDK components on project import ^[12].

The rest of the release follows the same idea. JetBrains describes the Toolchain as a unified entry point with a declarative configuration for JVM, Android, iOS, multiplatform and server-side builds ^[1]. It is one CLI for configuring the environment and for creating, building, testing and publishing KMP apps and libraries ^[5]. The new download progress bars show how far that goes. The Toolchain fetches JDKs, Android SDK components, iOS runtimes and compilers itself ^[11]. JetBrains is clear about who this is for: "Android developers might be discovering iOS for the first time," the company wrote ^[14].

The recommendation covers "most new" KMP projects ^[2]. It comes on a 0.x release of a tool JetBrains promoted to Alpha earlier this year ^[3]. Recent releases added web and cinterop support ^[6]. JetBrains wrote that "Our internal dogfooding and positive community feedback have given us the confidence to recommend it for new projects" ^[13]. I think it is the right place to start for a greenfield app built by an Android team adding iOS, because environment setup is the first cost that team pays. A team with an established build and a fixed release date is in a different position. For that team the choice is a 0.x Alpha tool, and the recommendation itself is limited to new projects ^[2].

The agent claim needs separate treatment. JetBrains says that in its benchmarks, AI agents generate and maintain Toolchain projects more reliably while consuming fewer tokens ^[16]. The post does not publish the tasks, models or scores behind that. JetBrains credits automatic environment setup, declarative configuration and clear, actionable error messages ^[15]. The result would carry over to your team only if your agents spend most of their time on the work the benchmark covered, which was generating and maintaining project builds ^[16]. Agent time spent on application code is outside what the benchmark measured.