zfinix/eleven is a reverse engineering tool that lets agents read and rename code or propose patches in macOS, Windows and Linux binaries, Android apps and GPU kernels. The latest release in the GitHub repository is v0.1.9. The provided information does not separately state what changed in this version.
The key point is that it connects agents supporting the Agent Client Protocol (ACP) so they can help with analysis in the same workbench. Native code is read through Ghidra-based REA or Rizin, and Android packages are analyzed with jadx. Agents follow code and references to propose function names, leave notes, or suggest patches. CUDA and HIP fat binaries can be opened as PTX, with support for rewriting them in CUDA C++, HIP C++ or Metal. It also handles functions and pseudocode for Mach-O, ELF and PE, along with assembly, control flow graphs and cross-references. Metal shaders and GGUF, safetensors and ONNX model files open in the same workbench. Changes keep an author record and can be reverted.
What this configuration means is bundling different analysis engines into an agent-centric workflow. That said, the product itself is close to a UI layer connecting existing engines, while deep analysis is handled by Ghidra, Rizin and jadx. So it is closer to a workbench for using several tools together with an agent than an independent analysis engine. The specific changes in v0.1.9 cannot be confirmed from the provided information alone.