The biggest improvement in this release is the automated generation of instructions for AI agents.
For years, one of the biggest challenges for JXRay users — especially those working on large, unfamiliar, or legacy applications — has been turning JXRay findings into actual code changes. Identifying the code responsible for issues detected by JXRay, such as duplicate strings or underutilized collections, and determining how to optimize it safely, can require significant time and effort. Less experienced developers frequently struggle to know where to begin and may abandon the effort altogether.
The good news is that advanced AI models, such as Claude Code Opus 4.6+, have recently become much better at analyzing JXRay reports and implementing practical, working optimizations. In some cases, their performance can be surprisingly effective. However, AI models do not always excel at selecting a small set of optimizations that are simultaneously high-impact, simple to implement, and safe.
To address this, JXRay now generates AI-focused optimization instructions directly within the report (available in paid mode). These instructions help guide the AI toward the most valuable opportunities identified by JXRay. To use them, simply copy the instructions into an .md file and provide it to your AI agent, or ask the agent to read the instructions directly from the beginning of the report.
Other changes in this version:
- Improved analysis of non-typical OOM heap dumps (containing a lot of garbage, smaller than maximum heap size, etc.)
- Reporting new types of potential memory leaks, for example small number of big objects taking most memory
- Detecting more kinds of JVM misconfiguration
- Bug fixes and miscellaneous improvements