PocketRecon-AI
hXR16F/PocketRecon-AI
Reconhecimento de segurança leve para IA em Raspberry Pi e PCs locais
Ver no GitHubMCP 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-AI1 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
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.
| Feature | PocketRecon-AI | mcp-kali-server | PentAGI | HexStrike-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:
| Component | Role | File |
|---|---|---|
| MCP Server (client) | MCP-compatible server, communicates with LLM | pocket_recon_mcp.py |
| Backend API (server) | Flask server that executes commands | pocket_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):
curlpingtraceroutenetcatnmap
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
- Open LM Studio and edit
mcp.json. - 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:
| Tool | Description | Parameters | API Endpoint |
|---|---|---|---|
ping | ICMP echo requests for reachability testing | target, count | POST /api/ping |
nmap | Network 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_args | POST /api/nmap |
netcat | TCP connectivity and raw socket interaction | target, port, data | POST /api/netcat |
traceroute | Trace network route to destination | target | POST /api/traceroute |
whois | Retrieve domain registration data | domain | POST /api/whois |
curl | Perform customizable HTTP requests using curl | url, method, data, json_body, headers, query_params, cookies, follow_redirects, timeout, connect_timeout, verbose, extra_args | POST /api/curl |
dirbuster | Web directory brute-force enumeration | target | POST /api/dirbuster |
http_request | Advanced HTTP interaction with redirect control | method, url, headers, params, data, json, follow_redirects, max_redirects | POST /api/http_request |
whatweb | Technology fingerprinting (CMS, frameworks) | target | POST /api/whatweb |
dns_enum | DNS record analysis and subdomain discovery | domain | POST /api/dns_enum |
header_security_audit | Security header audit (HSTS, CSP, etc.) | url, method | POST /api/header_security_audit |
cve_lookup | Passive CVE lookup via NVD keyword search | product, version | POST /api/cve_lookup |
abuseipdb_checker | IP reputation check using AbuseIPDB | ip | POST /api/abuseipdb_checker |
visit_website | Extract structured content from webpage | url, max_links, content_limit, timeout | POST /api/visit_website |
web_search | Lightweight web search using DuckDuckGo | query, limit | POST /api/web_search |
Security Best Practices
- Network Isolation: Run PocketRecon-AI in an isolated network segment
- Access Control: Restrict backend API to localhost or trusted IPs
- Logging: Enable audit logging for all tool executions
- 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 💙