Skip to main content
Glama

rea-mcp

MCP server wrapping the rea-agents CLI (REA: Reverse Engineer Anything) using FastMCP.

Every rea CLI command is exposed as an MCP tool. Tool names, descriptions and JSON Schemas are generated dynamically from rea --llms and rea <command> --schema, so a rea release upgrade adds tools without any code change here.

Prerequisites

  • Python 3.10+

  • Node.js 22.19+ (or 24.11+ / 26+) — required by the rea CLI

  • The rea CLI: npm i -g rea-agents

The server resolves the CLI in this order:

  1. REA_MCP_CLI environment variable (path to rea/rea-agents executable or the package's scripts/rea.mjs)

  2. rea / rea-agents on PATH

  3. A cached npx install under the npm cache (npm_config_cache or ~/.npm)

Related MCP server: pyghidra-mcp

Install

pip install -e .

Run

# stdio (for MCP clients)
rea-mcp

# HTTP
rea-mcp --transport http --host 0.0.0.0 --port 8000

Tools

One tool per rea CLI command (145+ in rea 6.5.0), named rea_<command> with dashes replaced by underscores, e.g.:

  • rea_analyze — app overview

  • rea_analyze_javascript_application — statically reconstruct a JS/Electron app

  • rea_inspect_go_binary / rea_inspect_rust_binary — embedded build metadata

  • rea_search / rea_function / rea_decompile / rea_xrefs — native analysis

  • rea_inspect_sqlite_database — SQLite snapshots

  • rea_inspect_managed_artifact / rea_inspect_managed_members — .NET assemblies

  • rea_inspect_wasm_artifact / rea_inspect_evm_interface — WASM / EVM

  • ADB/Android, browser observation, packet captures, Evidence import/export/compare …

Arguments map directly to the CLI: positional arguments keep their names, and camelCase options become kebab-case flags (e.g. caseSensitive: true → --case-sensitive, sourcePath: ["a.js"] → repeated --source-path a.js). All tools return the command's JSON output.

Provider-dependent commands

Native deep-analysis commands (analyze, search, decompile, …) require an analysis engine: Ghidra (GHIDRA_INSTALL_DIR), Hopper, or IDA (REA_IDA_MCP_CONFIG). Commands that need no engine (inspect-go-binary, inspect-sqlite-database, inspect-managed-*, JavaScript reconstruction, …) work out of the box. Run rea doctor through your MCP client, or check rea providers, to see what is available.

Skipped commands

setup, uninstall, update, completions and mcp are lifecycle/integration commands and are intentionally not exposed as tools.

License

MIT

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    D
    maintenance
    Enables LLMs to autonomously reverse engineer applications by exposing Ghidra's core functionality through MCP tools. Supports decompiling binaries, analyzing code structure, and automatically renaming methods and data.
    2
    Apache 2.0
  • A
    license
    Not graded
    quality
    B
    maintenance
    Exposes Ghidra reverse engineering capabilities via MCP, enabling LLMs and agents to analyze binaries, decompile, search, and edit programs headlessly or with GUI integration.
    454
    Apache 2.0
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables LLMs to autonomously reverse engineer applications by exposing Ghidra decompilation, renaming, and analysis tools to MCP clients.
    Apache 2.0
  • A
    license
    A
    quality
    C
    maintenance
    Publishes four flat, schema-light tools that let AI agents (including strict providers) search, describe, and invoke any of 139 reverse-engineering capabilities behind a persistent REA server. Enables file fingerprinting, inspection, and evidence-linked investigation of shipped software through plain-English requests, read-only by default.
    4
    62 npm
    1
    MIT