Databricks Open-Sources Metals v2

Databricks has open-sourced Metals v2, a significant upgrade to its Java and Scala language server, specifically engineered to tackle the challenges of multi-million line codebases. This move aims to enhance code intelligence for developers working within massive monorepos, a common scenario in large tech organizations. You can read the full announcement on the Databricks blog.

The update addresses a growing need for more efficient developer tools as AI coding assistants become more prevalent. With AI agents now generating a substantial portion of code, the emphasis shifts from exhaustive autocomplete features to rapid setup, reliable codebase orientation, and fast feedback loops. Metals v2 is designed to deliver precisely that, making it a vital component for developers who still engage in hands-on coding, especially within complex Java and Scala projects.

The IDE Flywheel in Action

Databricks standardized its internal development workflow around the Cursor IDE in May 2025, extending its use beyond frontend and non-JVM work. The primary hurdle for this consolidation was supporting their extensive Scala and Java monorepos, which house millions of lines of code. Metals v2 emerged as the solution, enabling Cursor and other lightweight editors like VS Code and Neovim to provide robust code navigation for these large JVM projects. This unified IDE approach creates a powerful flywheel effect: shared tooling improves developer collaboration and allows the platform team to concentrate investment, further enhancing the developer experience.

The impact of this shift is quantifiable. By July 2026, 92% of weekly active IDE users at Databricks were using Cursor, a dramatic increase from before the Metals v2 rollout. Similarly, Cursor's share of Scala and Java file open events climbed from 40% to 78% after the initial Metals v2 improvements landed in October 2025. This adoption trend extends beyond Databricks, with companies like Stripe already piloting Metals v2 for their extensive Java codebases, validating its scalability and performance.

Rethinking Code Intelligence for Scale

Metals v2 fundamentally rethinks how language servers operate for massive code repositories. Traditional language servers often struggle with the sheer volume of code, leading to slow startup times and unreliable code intelligence. Metals v2 tackles this by reworking three core layers: the repository index, the compiler pipelines, and the build integration boundary.

The "build-free repo index" uses an `mbt` index to scan workspace sources directly, bypassing the Build Server Protocol (BSP) on startup. This allows for rapid retrieval of essential information like cross-file references and jump-to-definition, even before a full build sync. This layer provides repo-wide intelligence, answering critical "first mile" questions quickly. For instance, fuzzy symbol searches across 2.9 million workspace symbols take only milliseconds.

The compiler pipelines for Scala and Java are built around the presentation compiler, a mode of the Scala type-checker that caches and reuses symbol table information. This allows a single instance to keep a 24-million-line Scala codebase in scope, publishing diagnostics with sub-second latency. After a build sync, a more precise compiler uses build server-reported classpath and sourcepath boundaries for greater accuracy.

Why This Matters for Developers and Enterprises

The open-sourcing of Metals v2 by Databricks, a company valued at $134 billion according to StartupHub.ai data, signifies a critical evolution in developer tooling. As AI coding assistants become more integrated into workflows, the underlying infrastructure supporting these tools must adapt. Metals v2's focus on speed, reliability, and scale directly addresses the pain points encountered when working with large, complex codebases that were previously only adequately supported by heavyweight IDEs like IntelliJ. This democratizes high-performance code intelligence, making it accessible in lightweight editors that are often preferred for their quick startup and low resource consumption.

This development is particularly relevant for startups and enterprises grappling with technical debt and the need for efficient developer productivity. For organizations like Stripe, which has also adopted Metals v2, it means a more streamlined development process and happier engineers. The collaboration between Databricks, the upstream Metals team, and Cursor highlights a growing trend of open-source collaboration driven by specific, large-scale enterprise needs. This model ensures that the tools developed for massive monorepos can benefit the broader JVM community, fostering innovation in developer productivity.

While Databricks is a titan in the data and AI platform space, with a StartupHub score of 82/100 and verified financials showing $7B raised, its competitors also offer compelling developer experiences. Companies like Alphabet Inc. (NASDAQ:GOOGL) (score 79/100) and Palantir (NASDAQ:PLTR) (score 85/100) are also deeply invested in developer tools and AI integrations. Cursor, with its StartupHub score of 77/100, is a direct beneficiary and early adopter, demonstrating how specialized tools can enhance even established platforms.

The future of coding is increasingly hybrid, blending human expertise with AI assistance. Tools like Metals v2 are foundational to this future, ensuring that developers can navigate and contribute to even the most complex software systems efficiently and effectively, regardless of their editor of choice.