JetBrains MPS 2026.2 Release Candidate Introduces ARM64 Support and AI Enhancements
The latest MPS release candidate improves Linux ARM64 compatibility and optimizes AI agent performance.
1Release Candidate Availability
JetBrains has officially released the MPS 2026.2 Release Candidate, allowing developers to test upcoming features before the final launch. This version includes numerous bug fixes and critical improvements over the initial EAP1 build. Users are encouraged to download the build now to verify compatibility within their existing development environments and workflows.
2Linux ARM64 Platform Support
MPS 2026.2 now provides native support for Linux on ARM64 architectures, including a bundled JetBrains Java Runtime. Developers can generate standalone MPS-based applications for this platform using the updated Build Script Wizard. This expansion ensures that native performance is accessible across modern ARM-based hardware configurations for all project distributions.
3Advanced TAR Archive Handling
The Build Language now supports importing archive contents directly into distribution layouts using new file import commands. This update preserves symbolic links and file permissions, which is essential for packaging native libraries correctly. By moving macOS distributions to .tar.gz format, the system prevents duplicate library inclusions that previously occurred with zip files.
4AI Agent Performance Boost
The Projectional Agent Toolkit has undergone a substantial rework of the MPS MCP API, resulting in over 50% token savings for common tasks. By utilizing compact summaries and lightweight metadata views, AI agents can now perform DSL modeling and language definition tasks with significantly reduced overhead and improved response accuracy.
5New Typesystem MCP Tool
A dedicated Typesystem MCP tool has been introduced to help agents interpret node types and debug complex rules. Agents can now trace type calculations, identify the specific rule responsible for errors, and determine expected types for roles. This integration simplifies the debugging process by providing agents with direct access to internal logic.
6Expanded Agent Capabilities
Beyond typesystem support, new MCP tools allow agents to invoke intentions, filter logs, and manage project lifecycles. These tools enable automated updates to skill databases and provide assistance in authoring concept functions. By correcting SModel expressions in parsed Java statements, agents can now handle more complex development tasks with higher precision.
7Import Dialog Functional Overhaul
The Import dialog has been redesigned to improve usability when selecting models and languages. Users can now utilize checkboxes for individual entries or bulk selection, fixing previous issues where selections were ignored. The interface now correctly reflects the state of imported components, ensuring that only the intended models are included during code pasting.
8Diff Dialog Visual Consistency
The Diff dialog has received UI updates to align with the broader platform aesthetic, including improved icon usage and vertical centering. Technical fixes ensure that change markers remain correctly positioned during resizing and that title bars stay aligned. These refinements provide a more stable and predictable experience when reviewing model changes.