MRPS
← Back to Toolkit

REA

MCP server

Model Context Protocol server exposing tools to AI agents via standard protocol.

1. What it does

An MCP server enabling agents to reverse engineer compiled applications, native binaries, Electron apps, Android APKs, and runtime behavior with deep inspection and decompilation capabilities.

2. Why MRPS selected it

Provides deep binary, package, and runtime reverse engineering capabilities through standardized Model Context Protocol tools.

3. Tool Kind

MCP server: Model Context Protocol server exposing tools to AI agents via standard protocol.

4. Original repository

https://github.com/morluto/rea

5. Official documentation

https://github.com/morluto/rea/blob/main/docs/mcp-contracts.md

6. Author and license

Created by Morluto under the MIT license.

7. Installation

npx rea-agents setup

8. When the AI should use it

Use when analyzing compiled binaries, APKs, Electron apps, or closed-source runtime behavior where source code is unavailable.

9. When the AI should not use it

Do not use for ordinary source-code architecture reviews when clean source code is already present.

10. What the AI must not do

  • Before you invoke anything, state what you intend to do and why. Do not act until the user confirms. If a rule is ambiguous, ask — do not guess.
  • Restrict analysis to explicitly provided target files or processes.
  • Never modify or patch target binaries during analysis.
  • Preserve evidence trails and distinguish between observed factual behavior and decompiled inferences.

11. MRPS usage instructions

Read REVERSE-ENGINEERING.md

12. Last verified

Verified against live repository on: 2026-11-07

View Complete Agent Constraint Sheet
Tool: REA When to use: Use when analyzing compiled binaries, APKs, Electron apps, or closed-source runtime behavior where source code is unavailable. When NOT to use: Do not use for ordinary source-code architecture reviews when clean source code is already present. Invocation: Register REA as an MCP server using npx -y rea-agents@latest. Rules: - Before you invoke anything, state what you intend to do and why. Do not act until the user confirms. If a rule is ambiguous, ask — do not guess. - Restrict analysis to explicitly provided target files or processes. - Never modify or patch target binaries during analysis. - Preserve evidence trails and distinguish between observed factual behavior and decompiled inferences.