Reverse Engineer Anything (REA):
Set Up Your First Analysis
REA connects AI coding agents to tools for inspecting software. This independent guide helps you configure REA for Claude Code, Codex or Cursor, choose the right analysis path, and check your results against reproducible evidence. Start with a setup plan for your environment.
$ npx rea-agents@6.5.0 setupFind the entry point. Show the evidence.
Illustration based on our recorded source analysis
Set Up Reverse Engineer Anything for Your Environment
Choose your OS, agent and target to generate steps and copy commands here. This browser-local assistant does not inspect or change your computer. Commands pin REA 6.5.0; generating a plan does not verify an installation.
Start with your environment
Run node --version and prepare a supported runtime before connecting your agent.
node --version
npm --version
npx rea-agents@6.5.0 setup --client claude_code --dry-run
npx rea-agents@6.5.0 setup --client claude_code
npx rea-agents@6.5.0 doctor --client claude_code --jsonPreview the plan, approve setup in your terminal, and restart Claude Code. Doctor verifies registration; check a small analysis target afterward.
A Reproducible Reverse Engineer Anything Case Study
We used REA 6.5.0 to inspect a frozen version of this site’s JavaScript setup wizard. The downloadable case includes the input files and recorded JSON output, so you can compare the findings with the source. It also documents the macOS SwiftBridging error encountered during the run and the project-local workaround used to complete the analysis.
These results belong to the frozen input in the case kit, not a new analysis of the current homepage. They do not establish compatibility with every platform or engine. Zero parse failures does not mean every inference is correct: check source locations and resolved versus unresolved relationships.
One question. A clear next step.
Choose the guide that gets you moving.
Connect your agent
Choose an OS, agent and target. Get a focused checklist and copyable setup commands.
Open the setup assistant ↗Inspect a working example
Trace a frozen version of this site’s wizard. Download the source and analysis results to reproduce it.
Read the case study ↗Find the failing step
Narrow down runtime, MCP registration, engine and Swift compiler issues.
Find a troubleshooting path ↗A bridge between your agent
and the software you inspect.
REA stands for Reverse Engineer Anything. It exposes reverse-engineering capabilities to coding agents through MCP. JavaScript, browser and native-binary investigations have different requirements; a successful setup is only the beginning.
Start with one feature, such as how a setup command is generated. Tie each conclusion to a source location or an observation you can check.
See our evidence →Looking for the REA GitHub project?
The upstream project is morluto/rea; its npm package is rea-agents. If you searched for “github rea”, “rea github” or “reverse engineer everything github”, check the project identity: the official name is Reverse Engineer Anything. rea.best is an independent guide.
An MCP connection is not an analysis engine.
MCP connects an agent to tools. REA coordinates analysis capabilities; Ghidra is a native-program analysis engine. Projects called Ghidra MCP may be separate bridges with different setup requirements. Their installation commands are not interchangeable.
You do not need Ghidra for the static JavaScript case here. Prepare an engine when your target actually needs native-binary analysis.
Choose an analysis path →Reverse Engineer Anything: Four Questions Before You Start
What is Reverse Engineer Anything?
Reverse Engineer Anything, or REA, connects AI coding agents to software-analysis tools. It provides different investigation paths with dependencies that vary by target. “Reverse engineering anything” is also a related search phrase; check the official morluto/rea project. This website is an independent guide.
Do I need Ghidra to use REA?
No. The static JavaScript case on this site does not require Ghidra. For native decompilation, prepare a compatible Ghidra, Hopper or IDA installation. Browser investigations use a suitable browser debugging session. Choose the target first, then prepare its dependencies.
Does REA recover the original source code?
There is no general guarantee. Native decompilation typically produces pseudocode, not the original source, variable names, comments or build environment. With readable JavaScript, analysis can inspect the actual supplied source. That is different from recovering source from a binary.
How do I know my first analysis succeeded?
Verify three things: your client discovers REA tools; an analysis runs on a small target whose behavior you understand; and its cited locations, relationships or observations match the input. A passing doctor check does not establish successful target analysis. Keep the version, environment and output, and label remaining inferences.
Investigating claims about Adobe? Check the evidence and capability limits →
Make your first investigation a small one.
Connect the tools, then follow a feature you care about.