REA, short for Reverse Engineer Anything, integrates AI coding agents like Claude Code and Cursor with inspection tools that work without source code This article explores agents reverse engineering. . This open-source project allows Ghidra, IDA Pro, and Hopper to be used in an agent-driven workflow, helping researchers trace program behavior and provide evidence to explain findings.

Its capabilities extend to native binaries, JavaScript, Electron applications, .NET assemblies, Android packages, firmware, websites, and selected runtime activity. REA Integrates Agents with Reverse Engineering Tools REA employs the Model Context Protocol (MCP) to enable agents to request analysis through a local server. The agent initiates the process by requesting tools to inspect a target, following relevant code, and receiving findings with evidence and stated limitations.

The official REA repository offers installation instructions and notes, detailing platform support based on the chosen analysis provider. In the DX-Ball case study, a recovered sound positioning calculation passed 3,205 tests against the original x86 code, reproducing all 63 compiled function bytes. Another example traces Notion’s clipboard handling through Electron’s renderer, preload layer, and main process.

Those results remain subject to the model provider’s data policy, so local execution does not guarantee that all findings stay on the machine. Its value lies in connecting questions to inspectable code and testable findings, while maintaining evidence, authorization, and human review at the forefront.