Made a project to integrate GPT models into directly Ghidra (github.com)

GhidraGPT is an open-source Ghidra plugin that embeds LLMs directly into reverse-engineering workflows to provide AI-powered code explanation, automated function/variable renaming, and vulnerability/security analysis. The plugin exposes these capabilities via a right‑click context menu and a dedicated console, integrates with Ghidra’s analysis pipeline, supports real‑time streamed responses for interactive use, and stores API keys encrypted through a user-friendly configuration panel. It’s buildable via the included Gradle scripts (./build.sh, ./install.sh) and targets Ghidra 10.0+ with Java 17. Technically notable is multi‑LLM support—OpenAI, Anthropic, Google Gemini, Cohere, Mistral, DeepSeek, Grok (xAI) and local Ollama models—handled by a modular service layer (GPTService, CodeEnhancementService, CodeAnalysisService). For low latency interactive work the author recommends smaller/faster models such as grok-3 or deepseek-chat; larger reasoning models will increase response time. Planned/available additions include variable retyping, cross‑reference analysis, batch processing, caching, custom prompts and export of results. Except for Ollama (local), providers require internet access and valid API keys. Significance: by tightly integrating LLMs into Ghidra, GhidraGPT can speed triage, improve readability, and surface potential vulnerabilities, making it a practical assistive tool for reverse engineers and security researchers—while reminding users that analysis quality and latency depend on model choice and code complexity.