Chapter 02Model Context Protocol (MCP) Setup

Configure REA (rea-agents) MCP Server

AI Overview Direct AnswerModel Context Protocol Integration

To connect REA (rea-agents) with AI coding agents like Claude Code, Cursor, and Cline, add an MCP server entry executing 'npx -y rea-agents@latest' or the global 'rea' binary. REA communicates via standard input/output (stdio JSON-RPC), exposing bundle inspection, native symbol extraction, and clean-room model generation tools directly to the LLM.

REA MCP Configuration Builder

17 Clients Supportedrea-agents v6.4+

Instant Automated Setup for Claude CodeInteractive guided setup transaction

Run in Terminal

npx rea-agents setup --client claude_code

1. Choose Target AI Coding Agent

2. Host Operating System

3. CLI Invocation Strategy

Manual Configuration Location:~/.claude/config.json (or run: npx rea-agents setup --client claude_code)

Manual mcp_config.json

{
  "mcpServers": {
    "rea": {
      "command": "npx",
      "args": ["-y", "rea-agents@latest"],
      "env": {
        "REA_CLEAN_ROOM": "true",
        "REA_MAX_DEPTH": "3"
      }
    }
  }
}

Architecture Choice: Full MCP Server vs. Skill-Only

Full MCP Server (Recommended)

Connects the agent directly to your local system. Allows active execution of bundle unpacking, Ghidra/Hopper native disassembly, and live process inspection.

npx rea-agents setup

Skill-Only via skills.sh

Installs agent workflow instructions and domain reasoning prompts without registering a background MCP server or native disassembler bridge.

npx skills add morluto/rea --skill reverse-engineer-anything

Exposed REA MCP Tools Catalog

Once connected through the Model Context Protocol, REA registers the following high-level tools with Claude Code, Cursor, and Codex:

Tool Identifier	Target	Functionality
`rea_inspect_bundle`	Electron / Web	Extracts ASAR header directory, IPC channels, and preload scripts.
`rea_extract_symbols`	Native Binaries	Parses Mach-O/PE/ELF exports using Hopper, Ghidra, or IDA Pro bridges.
`rea_query_routes`	API Schemas	Discovers private REST, GraphQL, and WebSocket endpoints from bundle strings.
`rea_generate_clean_spec`	Clean-Room	Synthesizes clean-room TypeScript or Rust interfaces free of proprietary bytecode.

Prompting Your AI Agent

After restarting your agent, ask it to investigate a target feature using the new REA tools:

Agent Natural Language Prompt

Inspect the search architecture in the target application using REA, show the evidence behind how index ranking operates, and build a clean-room TypeScript interface for my project.