cyberkaida/reverse-engineering-assistant

What it solves

ReVa addresses the difficulty of using AI for complex reverse engineering tasks, such as analyzing large binaries or firmware. It solves the problem of "context rot"—where AI models lose track of relevant information during long analysis sessions—by providing a structured, tool-driven interface that keeps the AI focused on critical, manageable fragments of data.

How it works

ReVa acts as a bridge between the Ghidra reverse engineering platform and AI language models using the Model Context Protocol (MCP). Instead of dumping massive amounts of data into the AI, ReVa provides a suite of specialized, small tools that the AI can call on demand. These tools supply specific, relevant information like namespaces, cross-references, and decompilation snippets. This approach mimics how a human reverse engineer uses a toolkit, nudging the AI to explore the binary methodically while reducing hallucinations.

Who it’s for

It is designed for security researchers, CTF participants, and developers who use Ghidra and want to integrate AI assistance into their workflow. It supports both interactive analysis (via the Ghidra UI) and automated, headless analysis for CI/CD pipelines or batch processing.

Highlights

• Tool-Driven Design: Provides the AI with small, specific tools to reduce context overload and improve accuracy.
• Interactive & Headless Modes: Works alongside the Ghidra GUI for deep, real-time analysis or runs in the background for automated tasks.
• MCP Compatibility: Integrates with standard MCP clients like Claude Code and VSCode, allowing it to be combined with other tools like GitHub or web search.
• Human-like Exploration: Includes context nudges (like cross-references) to encourage the AI to investigate binaries similarly to a human expert.
• Extensible: Supports custom skills for specific tasks like binary triage, cryptography analysis, and CTF challenges.