Build my stack
Major Release
JetBrains Blog: IntelliJ IDEA Goes LSP
Aug 4, 2026
- JetBrains released a preview VS Code extension that brings IntelliJ IDEA's Java and Kotlin language intelligence via LSP to VS Code, Cursor, and compatible editors.
- Includes smart completion, navigation, refactoring, and full support for Maven, Gradle, and Bazel projects, even in large monorepos.
- Internal trials show the LSP significantly improves terminal-based agentic workflows (Claude Code, Codex) by reducing token consumption for code understanding.
- Free during preview (30-day builds). Will require IntelliJ IDEA Ultimate license after stable 1.0 release.
JetBrains announced on August 4, 2026, that IntelliJ IDEA's Java and Kotlin code intelligence is now available as a preview LSP extension for VS Code and its forks, including Cursor. This is the first time JetBrains has officially shipped its language analysis engine outside of its own IDEs.
The strategic context is obvious: developers are increasingly working in VS Code and AI-native editors like Cursor. Rather than watch those users get inferior Java tooling, JetBrains is meeting them where they are while keeping IntelliJ IDEA Ultimate as the premium tier.
What the Extension Delivers
The extension brings smart code completion, navigation, code analysis, and refactoring for Java, Kotlin, and mixed-language projects. It supports Maven, Gradle, and Bazel build systems and handles large monorepos without choking, a common pain point with existing VS Code Java extensions.
This isn't a lightweight linter. It's the same analysis engine that powers IntelliJ IDEA, packaged as an LSP server. The quality gap between JetBrains' code understanding and existing open-source Java LSP implementations (Eclipse JDT-LS) has been a recurring complaint from developers who switched to VS Code. That gap just closed.
The AI Agent Play
The more interesting angle is agentic workflows. JetBrains revealed that internal trials with Claude Code and Codex showed the LSP significantly reduces token consumption when AI agents need to understand codebases. Instead of reading thousands of files to infer types and relationships, agents can query the LSP for precise answers.
JetBrains is working on dedicated agent plugins that expose refactoring, test generation, and project structure through tool-call interfaces. The thesis: AI coding agents shouldn't reconstruct project knowledge from raw files. They should use the same deterministic analysis that IDEs have refined over 20 years.
Pricing and Availability
During preview, the extension is free with 30-day time-limited builds. After stable 1.0, it will require an IntelliJ IDEA Ultimate license ($599/year individual, $149/year with continuity discount). This positions it as a premium upsell: free editors get free extension preview, but long-term use funds ongoing JetBrains development.
Install it today from the VS Code Marketplace or Open VSX registry. Open any Maven, Gradle, or Bazel Java/Kotlin project and the extension auto-imports and indexes it.
Frequently Asked Questions
Does this work in Cursor?
Yes. Cursor is a VS Code fork, so the extension installs and works identically. Java/Kotlin code intelligence is immediately available in Cursor's editor and AI agent context.
Will it be free forever?
No. Free during preview (30-day builds). After 1.0 stable release, it requires an IntelliJ IDEA Ultimate license.
How does it compare to the existing Java extension in VS Code?
The existing extension uses Eclipse's JDT-LS. JetBrains' extension uses IntelliJ's engine, which typically provides faster indexing, better refactoring support, and more accurate completions in complex codebases.
Does it support Spring Boot and Jakarta EE?
Yes. Framework-specific intelligence (Spring annotations, Jakarta EE APIs) works the same as in IntelliJ IDEA.
Can AI agents use this?
JetBrains is building agent plugins that expose the LSP capabilities through tool-call interfaces for Claude Code, Codex, and other terminal-based agents. Early results show reduced token usage.
Daniel Cho covers AI model releases, developer tools, and the frontier lab landscape. He has tracked the LLM space since GPT-3 and specializes in making model architecture understandable to practitioners.
