JetBrains Pivots to AI Agents, Unveils Junie Code Assistant with Free Tier and Local Execution
JetBrains, a long-standing leader in code editors and IDEs, has strategically entered the AI agent market with the introduction of Junie. Recognizing the industry’s shift away from purely manual coding, the company has positioned Junie as a comprehensive AI code assistant available across terminals, popular IDEs via the ‘Air’ plugin (WebStorm, GoLand, etc.), and directly within GitHub and GitLab workflows for issues and pull requests.
Key features of Junie include support for over ten different models, allowing users to bring their own API keys for advanced options like Claude Opus, Sonet, Fable, Gemini, and GPT-6 Astra. The cornerstone of its offering is ‘Junie Lite,’ a free tier leveraging a Guen 3.6 model (27 billion parameters), accessible with a JetBrains account—no credit card or trial required. This free access is guaranteed until December 31, 2026. Junie Lite is observed to be surprisingly fast and efficient for simpler tasks, such as generating changelogs from Git commits and handling basic frontend project requirements. It demonstrates ‘human-in-the-loop’ capabilities by asking clarifying questions and supports dynamic ‘live prompting’ to adjust its course during execution. However, during testing, Junie Lite encountered challenges with complex UI framework documentation (e.g., Astro components), requiring direct user intervention to rectify structural issues, and exhibited minor terminal rendering glitches.
For more demanding users, JetBrains offers paid AI Pro and AI Ultimate plans, alongside a compelling ‘Junie Local’ option. Junie Local enables on-premise model execution, providing unlimited tokens and offline functionality, but necessitates high-end hardware like an Apple Silicon M5 with 64GB RAM or an Nvidia RTX GPU with 24GB VRAM. This positions Junie Lite as an accessible entry point for developers to experiment with AI-assisted coding, while the paid and local options cater to those requiring more advanced capabilities and computational power.