reverse engineer anything

Reverse Engineer Anything (REA) is an open-source toolkit that connects AI coding agents to software inspection tools through the Model Context Protocol (MCP). It helps you investigate applications, understand how a feature works, and collect evidence from code and observed behavior. Its tools are also available through a command-line interface.

Your coding agent calls REA through MCP to inspect a target, locate relevant code, and trace relationships. REA returns findings with evidence and limitations. The agent uses those results to ask follow-up questions, explain behavior, or write and test an implementation. Native analysis uses a compatible engine such as Hopper, Ghidra, or IDA.

Install a supported Node.js version and npm, then run npx rea-agents@latest setup in your terminal. Select your coding agent, review the proposed configuration changes, and approve them. Restart the agent so it discovers the MCP server and workflow instructions. Configure any analysis engine required by your target; static JavaScript inspection can run without a native engine.

Start with one feature in software you are authorized to examine. Give your agent the target’s absolute file path or website URL and a focused question, such as where a CSV export is implemented. Ask it to use REA, trace the relevant code, and show supporting evidence. Review the findings, identify uncertainties, and test any implementation you create.

Yes. REA-authored code is open source under the MIT license, which permits use, modification, and redistribution subject to its notice requirements. Third-party components keep their own licenses. Your coding agent, model provider, or commercial analysis engines may have separate costs and usage terms; REA does not include licenses for those tools.

Supported targets include native executables and libraries, JavaScript and Electron apps, .NET assemblies, websites, and Android APKs. Depending on the workflow, REA can inspect functions, strings, modules, references, application resources, and selected runtime or network behavior. Platform support and required external tools vary by target, so check the official capabilities before starting.

Run npx rea-agents@latest setup and choose Claude Code or another supported client, such as Codex, Cursor, or Gemini CLI. Review and approve the MCP configuration, restart the client, and confirm that REA appears in its tool list. Then ask the agent to use REA on a specific target. Other clients that support local MCP servers can use manual registration.

REA can help an agent understand compiled or packaged software without its original source code and build an implementation of an observed feature. The result depends on the target and available evidence. A complete or identical reconstruction is not guaranteed, and recovered pseudocode is not the original source. Check the target’s rights and licenses before reusing or sharing code, assets, or findings.

Safety and legality depend on the target, permissions, and operations you request. REA is intended for authorized analysis, research, education, and interoperability. Review commands before execution, use an isolated environment for untrusted software, and check your agent provider’s data policy. Local analysis does not make REA a sandbox. Applicable laws, software licenses, privacy obligations, and terms of service still govern your use; possession of an app alone does not authorize every operation.

Choose tools around your target and workflow: Ghidra offers a free, open-source workspace for native analysis; IDA Pro and Binary Ninja offer specialist binary-analysis environments; Frida supports dynamic instrumentation; and Chrome DevTools helps inspect web applications. Ghidra and IDA can also work as REA analysis providers. Compare the supported platforms, licenses, and automation features that your investigation needs.