convez.ai
MCP ServerSegurançaAvançado
hXR16F

PocketRecon-AI

hXR16F/PocketRecon-AI

Reconhecimento de segurança leve para IA em Raspberry Pi e PCs locais

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server leve para reconhecimento de segurança cibernética alimentado por IA em dispositivos de baixa potência. Integra ferramentas como Nmap, Netcat e WhatWeb diretamente em interfaces conversacionais, permitindo que agentes de IA executem tarefas de reconnaissance de rede, análise web e coleta de inteligência de forma nativa.

Instalação

$npx convez-ai hXR16F/PocketRecon-AI

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automação de testes de penetração conversacionais
  • Reconhecimento de rede integrado em agentes de IA
  • Análise de segurança web assistida por LLM
  • Inteligência de ameaças passiva via busca de CVE e reputação de IP

Veredito

Ferramenta sólida e focada para profissionais de segurança que precisam integrar reconnaissance a fluxos de IA. Destaca-se pela leveza e compatibilidade multiplataforma, mas requer expertise em segurança ofensiva para uso responsável.

README

A tradução para o português deste README ainda está na fila.

📡 PocketRecon-AI


Overview

PocketRecon-AI is a Model Context Protocol (MCP) server that enables Large Language Models (LLMs) to perform cybersecurity reconnaissance tasks directly within conversational interfaces. Built for low-power devices like Raspberry Pi and local PCs. PocketRecon-AI connects AI agents with powerful security tools including Nmap, Netcat, WhatWeb, Curl, and more.

The system uses a two-tier architecture: an MCP server that communicates with your LLM (LM Studio, VS Code, Cursor, etc.), and a Flask-based backend API that executes the actual system commands.

PocketRecon-AI is cross-platform, and runs on both Linux and Windows environments.


Features

Network Reconnaissance

  • ICMP Ping – Reachability testing with configurable packet counts
  • Traceroute – Network path analysis to destinations
  • Nmap Scanning – Port scanning, OS detection, NSE script execution and more
  • Netcat – TCP connectivity testing and raw socket interaction

Web Analysis

  • Directory Brute-forcing – Web directory enumeration using dirb wordlist
  • Technology Fingerprinting – CMS and framework detection
  • HTTP Security Auditing – Header analysis (HSTS, CSP, etc.)
  • Web Content Extraction – Structured data extraction from webpages

Intelligence Gathering

  • WHOIS Lookup – Domain registration data retrieval
  • DNS Enumeration – Subdomain discovery and record analysis
  • CVE Lookup – Passive vulnerability search via NVD
  • IP Reputation – AbuseIPDB integration for threat intelligence
  • Web Search – Lightweight DuckDuckGo searching

LLM Integration

  • MCP Server – Native Model Context Protocol support
  • Works with: LM Studio, VS Code, Cursor, Ollama, and any MCP-compatible client
  • Automated security workflows via conversational interface

Technical Comparison

Comparison of AI-powered cybersecurity reconnaissance platforms, autonomous pentesting assistants, and MCP-based offensive security tooling.

FeaturePocketRecon-AImcp-kali-serverPentAGIHexStrike-AI
Weight🟢 Extremely Lightweight🟡 Moderate🔴 Very Heavy🟡 Moderate
Safety🟢 Abstracted wrapper-based execution🔴 Direct shell command execution🟢 Containerized sandbox execution🔴 Direct shell command execution
Dependencies🟢 Python + Nmap🟡 Kali Linux environment🔴 20+ GB stack + API keys🟠 150+ external tools
Setup Complexity🟢 Minimal🟡 Moderate🔴 High🟡 Moderate
System Requirements🟢 Runs on a Potato🟡 Low-end PC🔴 4+ GB RAM recommended🟡 Low-end PC
Capability Scope🟡 Reconnaissance & scanning🟢 Full Kali ecosystem🟢 Autonomous pentesting framework🟢 Broad offensive tooling

Architecture

PocketRecon-AI uses a two-component architecture that separates the LLM interface from the command execution:

ComponentRoleFile
MCP Server (client)MCP-compatible server, communicates with LLMpocket_recon_mcp.py
Backend API (server)Flask server that executes commandspocket_recon_server.py

The MCP Server (client) exposes tools to your LLM and forwards requests to the Backend API (server), which runs the actual security tools on your system.


Installation & Setup

Prerequisites

  • Python 3.10+
  • Docker (optional, for containerized deployment)
  • System Tools (must be installed on the backend API server):
    • curl
    • ping
    • traceroute
    • netcat
    • nmap

Option 1: Local Installation (Recommended for Raspberry Pi)

Step 1: Clone repository

git clone https://github.com/hXR16F/PocketRecon-AI
cd PocketRecon-AI

Step 2: Create virtual environment

python -m venv venv
source venv/bin/activate    # On Windows: venv\Scripts\activate

Step 3: Install dependencies

pip install -r requirements.txt
sudo apt install nmap    # On Windows: install from https://nmap.org/download#windows

Step 4: Run backend API Server

python pocket_recon_server.py --host 0.0.0.0 --port 5005

Option 2: Docker Deployment

Build Image

docker build -t pocket-recon .

Run Container

docker run -d \
  --name pocket-recon \
  -p 5005:5005 \
  --cap-add=NET_ADMIN \
  --cap-add=NET_RAW \
  pocket-recon

Connecting to your LLM

LM Studio

  1. Open LM Studio and edit mcp.json.
  2. Configure as follows:
{
  "mcpServers": {
    "PocketRecon-AI": {
      "command": "python",
      "args": [
        "/path/to/pocket_recon_mcp.py",
        "--server",
        "http://localhost:5005"
      ],
      "timeout": 900000
    }
  }
}

Available Tools

All tools are exposed to your LLM via the MCP protocol and map to backend API endpoints:

ToolDescriptionParametersAPI Endpoint
pingICMP echo requests for reachability testingtarget, countPOST /api/ping
nmapNetwork scanning (ports, OS detection, NSE scripts, etc.)target, ports, scan_type, scripts, os_detection, service_detection, script_scan, script_args, top_ports_count, timing, extra_argsPOST /api/nmap
netcatTCP connectivity and raw socket interactiontarget, port, dataPOST /api/netcat
tracerouteTrace network route to destinationtargetPOST /api/traceroute
whoisRetrieve domain registration datadomainPOST /api/whois
curlPerform customizable HTTP requests using curlurl, method, data, json_body, headers, query_params, cookies, follow_redirects, timeout, connect_timeout, verbose, extra_argsPOST /api/curl
dirbusterWeb directory brute-force enumerationtargetPOST /api/dirbuster
http_requestAdvanced HTTP interaction with redirect controlmethod, url, headers, params, data, json, follow_redirects, max_redirectsPOST /api/http_request
whatwebTechnology fingerprinting (CMS, frameworks)targetPOST /api/whatweb
dns_enumDNS record analysis and subdomain discoverydomainPOST /api/dns_enum
header_security_auditSecurity header audit (HSTS, CSP, etc.)url, methodPOST /api/header_security_audit
cve_lookupPassive CVE lookup via NVD keyword searchproduct, versionPOST /api/cve_lookup
abuseipdb_checkerIP reputation check using AbuseIPDBipPOST /api/abuseipdb_checker
visit_websiteExtract structured content from webpageurl, max_links, content_limit, timeoutPOST /api/visit_website
web_searchLightweight web search using DuckDuckGoquery, limitPOST /api/web_search

Security Best Practices

  1. Network Isolation: Run PocketRecon-AI in an isolated network segment
  2. Access Control: Restrict backend API to localhost or trusted IPs
  3. Logging: Enable audit logging for all tool executions
  4. Updates: Keep system tools (Nmap, etc.) updated regularly

Safety & Ethics Warning

FOR AUTHORIZED CYBERSECURITY PROFESSIONALS AND EDUCATIONAL USE ONLY

  • Unauthorized Scanning: Do NOT use this tool to scan networks, systems, or domains you do not own or have explicit written permission to test.
  • Legal Compliance: Ensure all usage complies with local laws and ethical guidelines.

Built with 💙

Ver o original no GitHub

Bom para

Relacionados