Databricks has open-sourced Metals v2, a Java and Scala language server designed to provide code intelligence across multi-million-line codebases and support lightweight IDEs.
Databricks has open-sourced Metals v2, an upgraded Java and Scala language server designed for large codebases. The software targets multi-million-line monorepos and aims to provide faster setup, reliable codebase orientation, and shorter feedback loops for developers working on complex Java and Scala projects.
Databricks implemented Cursor as its standard IDE development environment in May 2025 to address the challenges of working with large Scala and Java monorepos. By July 2026, 92 per cent of Databrick’s weekly active IDE users were using Cursor. Since the first major Metals V2 improvements were introduced in October 2025, Cursor’s share of Scala and Java file-open events had increased from 40 to 78 per cent.
Metals v2 reworks three major layers of the language server: the repository index, compiler pipelines and the build integration boundary. The repository index works without the build server during startup, using an mbt index to scan workspace sources directly. It offers users access to information like cross-file references, definitions etc before synchronisation. The index contains information on 2.9 million workspace symbols, with fuzzy symbol search completing at a median latency of 10 ms.
The Scala and Java compiler pipelines use compiler-based approaches to support code intelligence at a scale. The Scala pipeline uses a presentation compiler that caches and reuses symbol-table information across compilation runs. A single Scala compiler instance can keep the full 24-million-line Scala codebase in scope.
Metals v2 is designed to provide low-latency code intelligence for large JVM repositories through lightweight editors. Companies such as Stripe are currently experimenting with Metals v2 in handling large java repositories. The project is being developed for large Java and Scala codebases where conventional JVM language servers can struggle with scale.
