🔍 Read the full analysis: How To Start Investigating Existing Software With An AI Agent on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
ThorstenMeyerAI.com published a practical guide to using Reverse Engineer Anything (REA) with coding agents to investigate how existing software behaves. It recommends starting with one specific question, confirming the agent is connected to REA, and checking findings against concrete examples rather than trusting a plausible explanation.
ThorstenMeyerAI.com published a five-part guide on October 19 to using the open-source Reverse Engineer Anything (REA) toolkit with AI coding agents to investigate existing software. The guide centers on a practical question: how to establish what a program actually does—such as how it calculates an accounting balance—before using those rules to build or maintain another application.
REA connects coding agents, including Claude Code and Codex, to analysis tools through a local server, a command-line program and an agent skill. The project supports different kinds of targets, including native applications, JavaScript and Electron apps, .NET and Android software, databases, firmware and saved network traffic. For websites, it can observe activity in a Chrome-family browser controlled by the user. The toolkit is MIT-licensed and open source; the source article reported version 6.7.0 was released October 18.
The guide advises narrowing each investigation to one behavior and the decision it informs. Instead of asking an agent to “figure out” an entire accounting program, a developer might ask how a partial invoice payment affects the remaining balance and which account receives the posting. The agent should report stored data, calculations and conditions that change the result, while marking statements as observed, inferred or unknown.
Setup is not proof that the agent can use the tools. The article says REA’s guided installer can detect supported agents, while a diagnostic command checks files and prerequisites but does not confirm a live agent connection. Users should verify the connection from inside the agent and test it on software they own. A further check is whether the agent actually calls an REA tool: if it answers without doing so, its response may be a model-generated guess rather than an analysis result.
Infographic 01 · Reverse Engineer Anything
Five places to point REA first, and how to tell when the answer is wrong.
REA gives coding agents such as Claude Code and Codex tools to inspect programs, apps and browser sessions. Each investigation needs a finishable question and a check.
npx rea-agents@latest setupThe five investigations
What to ask, and the check that catches a wrong answer.
- 1
Ask a question you can finish
One behaviour, one decision it informs. Example: how a partial payment changes an invoice balance.
It’s wrong ifThe agent can't walk through one concrete example with real inputs and outputs. - 2
Get REA connected and confirmed
Run the guided setup, then doctor, then call binary_session inside the agent before any real target.
It’s wrong ifAnswers arrive without a single REA tool call: you're reading the model's guesses. - 3
Turn a business rule into a spec
Inputs, stored fields, formula, edge cases, plus test cases checked against the original's own output.
It’s wrong ifYour version and the original disagree on a test case, even by one cent. - 4
See what has to survive a copy
Find what an app writes to the clipboard and what it keeps locally, then design your own copy and paste.
It’s wrong ifThe explanation doesn't predict what lands in a plain text editor, an email and your app. - 5
Your own app, your equipment, an old program
Start with software you own or rely on. Ask what it sends, stores or calculates when you do one thing.
It’s wrong ifIt only works for the example you watched. Test with a new file or a read-only command first.
Sources: REA on GitHub (stars via GitHub API, 19 Oct 2026); REA installation guide; REA security policy.
Powered by Thorsten Meyer AI
From Software Behavior to Working Specs
The approach matters to developers who need to maintain older systems, replace a product or reproduce a business rule in a new application. A screen can show what a program displays without revealing what data it stores or how it reaches a result. Tracing a specific behavior can turn that hidden logic into a testable explanation and, potentially, a more useful implementation specification.
The article illustrates this with Ledger Reborn, a local accounting application whose author investigated QuickBooks Desktop for Mac 2024 using REA and Ghidra. According to the source article, the author tested the new application against Intuit’s official sample company and reported that its profit-and-loss statement, balance sheet and trial balance matched account by account, to the cent. That reported result is a case study, not independent proof that REA will produce equivalent results for other software or projects.
For readers using AI coding tools, the practical point is to treat the agent as an investigator whose work needs evidence—not as an authority. A focused question, a reproducible example and a record of what is known can help expose incorrect assumptions before they become features in a replacement system.
software reverse engineering tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How REA Connects to Coding Agents
REA is described in the source article as a local MCP server, a command-line tool named rea and an agent skill. MCP gives compatible agents a way to call external tools. The project’s analysis options include tools such as Ghidra, Hopper or IDA for native programs; native analysis may require additional software, including Java for Ghidra. The article lists Node.js 22.19 or newer as a requirement.
The guide’s QuickBooks example used REA 6.4.0 and Ghidra to examine a limited set of accounting behaviors: inventory arithmetic, budget sign changes by account type, reconciliation balances and cash-basis report paths. The source describes REA as supplying evidence for the investigation, not automatically producing a verified specification. The developer still has to check whether an explanation accounts for a complete example with real inputs and outputs.
One safety limitation also shapes the workflow: analysis runs locally, but results are sent to the coding agent and may be handled by its model provider under that provider’s data policy. The project warns, “This is not a sandbox.” Running an unknown program through analysis tools can expose the user’s machine to the same risks associated with handling other untrusted files.
“This is not a sandbox.”
— REA project, as quoted by ThorstenMeyerAI.com
As an affiliate, we earn on qualifying purchases.
Limits of the Reported Findings
The source article presents the QuickBooks comparison as the project author’s result; it does not provide an independent audit or enough detail here to establish how broadly the match was tested. The reported agreement applies to specific financial statements and a sample company, not every QuickBooks feature or accounting scenario. It is also unclear from the supplied material how often REA’s findings need manual correction across different software types.
REA’s local analysis does not mean all information stays local: findings go to the connected agent, and the source says the model provider’s data policy applies. The material does not specify the terms for every provider or explain how users should handle confidential source code. Those questions need to be checked against the relevant agent and provider policies before an investigation begins.
open-source software analysis toolkit
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Verify Tools Before Investigating
The guide’s next step for a user is to install REA with its guided setup, review the proposed changes and confirm prerequisites. After restarting the coding agent, the user should test that REA tools are available and perform an initial inspection on a harmless application they own. A diagnostic result alone is not enough; the agent must demonstrably call the analysis tool.
For the investigation itself, the recommended test is a complete, concrete example: give the program real inputs, trace the resulting output and ask the agent to separate observations from inferences and unknowns. The source material does not report a new REA release or a scheduled follow-up. It leaves developers with a repeatable method for checking claims before treating them as a specification.
software investigation tools for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is REA?
Reverse Engineer Anything is an open-source toolkit that connects AI coding agents to software-analysis tools through a local MCP server, command-line interface and agent skill.
Can REA prove that an AI agent’s explanation is correct?
No. The guide recommends checking whether the agent used REA and whether it can explain a complete example with real inputs and outputs. Findings may include inference or uncertainty and need to be evaluated.
Does REA keep all analysis data on my computer?
REA runs analysis locally, but its results go to the connected agent and may also go to the agent’s model provider. The applicable data handling depends on that provider’s policy.
What should I investigate first?
Choose one specific behavior tied to a decision, such as how a partial payment changes an invoice balance. Ask for the stored data, calculation and conditions that affect the result.
Is it safe to analyze an unknown program?
The project warns that REA is not a sandbox. The source article advises treating unknown binaries as untrusted files because analysis tools run with the user’s permissions.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
