MCPFP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@MCPFP Run a guided security assessment on the lab target and generate a Markdown report."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCPFP — Model Context Protocol For Pentesting
Servidor MCP (Model Context Protocol, SDK oficial de Python) para practicar evaluación de seguridad contra una máquina virtual de laboratorio (VulnHub u otra). Valida el OWASP Top 10, escanea red y protocolo, y guía al estudiante paso a paso.
Regla de oro: el MCP describe vulnerabilidades y sugiere acercamientos, pero nunca las explota. Los hallazgos se entregan en el chat y se guardan como reporte Markdown en el directorio desde donde se invocó el MCP. Úsalo solo en laboratorios propios o con autorización explícita.
Instalación
python -m venv .venv
.venv\Scripts\python -m pip install -e .
copy .env.example .envEdita .env con los datos del laboratorio (ver comentarios en .env.example):
MCPFP_TARGET_HOST: IP/host de la VM.MCPFP_AUTHORIZED=true: confirma que tienes permiso (obligatorio para operar).MCPFP_PORT_RANGE,MCPFP_REPORT_DIR, credenciales de lab (opcionales).
Nunca subas .env a un repositorio.
Herramientas externas recomendadas (el MCP valida su existencia y sugiere cómo
instalarlas): nmap, nikto, whatweb, openssl, curl, dig.
Related MCP server: mcp-security-server
Uso
Servidor MCP
Correrlo manualmente (stdio), invocando el módulo del servidor:
.venv\Scripts\python -m mcpfp.serverPara registrarlo en Claude Code:
claude mcp add mcpfp -s user -- "<ruta>\.venv\Scripts\python.exe" -m mcpfp.servero para otro cliente MCP (Claude Desktop, etc.), agrega este bloque a la config
de mcpServers del cliente, ajustando las rutas a tu máquina:
{
"mcpServers": {
"mcpfp": {
"command": "C:\\ruta\\a\\MCPFP\\.venv\\Scripts\\python.exe",
"args": ["-m", "mcpfp.server"],
"cwd": "C:\\ruta\\a\\MCPFP"
}
}
}cwd define dónde se escribe el reporte Markdown (por defecto en reports/).
Herramientas expuestas
Tool | Descripción | ¿Explota? |
| Carga y valida el objetivo del | No |
| Verifica herramientas externas y sugiere instalación | No |
| Puertos/servicios/versiones (nmap no intrusivo) | No |
| Cabeceras de seguridad HTTP | No |
| Certificado y protocolo TLS | No |
| Fingerprinting + mapeo OWASP Top 10 | No |
| Tutor: acercamientos obvios y no obvios | No |
| Guarda hallazgos en Markdown en el cwd | No |
Además, el prompt MCP metodologia describe el flujo completo.
Guardarraíles anti-explotación
src/mcpfp/guardrails.py aplica lista blanca de binarios y lista negra de
argumentos intrusivos (exploit, brute, dos, dump, etc.). Herramientas de
explotación (sqlmap, hydra, metasploit, …) están prohibidas.
Skill de Claude
skills/ciberseguridad-owasp-iso27001/SKILL.md: metodología de abordaje
defensivo siguiendo OWASP + ISO/IEC 27001.
Estructura
src/mcpfp/
├── server.py # FastMCP: tools, prompt
├── config.py # objetivo único desde .env
├── guardrails.py # anti-explotación
├── tools/ # environment, recon, protocol, web_owasp, guidance, report
└── knowledge/ # referencias OWASP / protocolo
skills/ # Skill de Claude (OWASP + ISO 27001)Aviso legal
Herramienta educativa y defensiva. El uso contra sistemas sin autorización es ilegal. El autor y los contribuyentes no se responsabilizan del mal uso.
Available Tools
8 toolsanalyze_http_headersC
Evalua las cabeceras de seguridad HTTP del objetivo.
Inputs: port, use_https. Returns: result
analyze_tlsB
Inspecciona el certificado y protocolo TLS del objetivo.
Inputs: port. Returns: result
assess_owaspB
Fingerprinting web y mapeo pasivo a OWASP Top 10.
Inputs: port, use_https. Returns: result
check_environmentA
Verifica que herramientas externas de reconocimiento estan instaladas.
Inputs: no named parameters. Returns: result
get_guidanceB
Guia pedagogica paso a paso (obvios / no obvios / siguiente paso).
Temas: reconocimiento, a01, a02, a05, a06, bandera.
Inputs: topic (required). Returns: result
load_target_infoA
Carga y valida el objetivo unico desde .env, confirmando autorizacion.
Inputs: no named parameters. Returns: result
scan_networkA
Descubre puertos y servicios del objetivo (nmap no intrusivo).
Inputs: no named parameters. Returns: result
write_reportB
Guarda hallazgos en un reporte Markdown en el directorio de invocacion.
Inputs: findings (required), append. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 8 tools
Each tool targets a distinct stage or concern: target setup, environment checks, network scanning, HTTP header analysis, TLS inspection, OWASP mapping, guidance, and reporting. Even the two analyze tools are clearly separated by layer, so no agent should confuse one tool for another.
All tool names follow a consistent snake_case verb_noun pattern: load_, check_, scan_, analyze_, assess_, get_, and write_. The verb clearly signals the action and the noun signals the target, making the naming predictable and easy to extend.
Eight tools is a well-scoped set for a reconnaissance and OWASP assessment workflow. Each tool has a clear responsibility and no tool feels redundant or extraneous.
The workflow is complete for the stated passive-assessment and pedagogical purpose: it covers target validation, environment readiness, network discovery, HTTP/TLS inspection, OWASP mapping, guidance, and report generation. There are no obvious dead ends or missing core operations.
Maintenance
Related MCP Connectors
Offline methodology engine for authorized penetration testing, CTF, and security research.
Scan a website for vulnerabilities: OWASP Top 10, CVEs, SSL, headers - with plain-English fixes
Audit active CISA KEV ransomware CVEs, SSL/TLS CT log subdomains, SPF/DMARC security & ICANN RDAP.
Passive OSINT and security recon: DNS, TLS, email posture, IP reputation, CVE/KEV, CT logs.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenancePassive security scanner that audits a running MCP server against the OWASP MCP Top 10 and grades it A-F. Read-only static analysis of the advertised tools, prompts and resources with console/JSON/SARIF output, and it also runs as an MCP server itself.MIT
- FlicenseNot gradedqualityDmaintenanceProvides AI agents with standardized security scanning capabilities for authorized assessments, including host detection, port scanning, service fingerprinting, web vulnerability scanning, CVE verification, and report generation, all with unified JSON output and strict compliance controls.-
- FlicenseAqualityCmaintenanceEnables controlled, evidence-first web application penetration testing and bug bounty workflows, including scope enforcement, reconnaissance, human-approved surface triage, OWASP WSTG test execution, and SHA-256 evidence capture.101-
- AlicenseBqualityBmaintenanceEnables read-only, evidence-led security audits of GitHub-hosted source code through explicit workflow primitives: repository reconnaissance, coverage-ledger planning, heuristic candidate leads, structured findings validation, independent verification gates, and JSON/Markdown report generation. It never clones, executes, or probes the target code, and marks runs incomplete until coverage and reviewer-verification gates pass.14AGPL 3.0