mcp-noctua
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., "@mcp-noctua scan example.com for vulnerabilities"
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.
mcp-noctua
🇬🇧 English · 🇫🇷 Français
An MCP server (stdio) that exposes a pentest toolbox to a strong LLM orchestrator (Claude Code, interactive), to run authorized security audits.
A clean rewrite inspired by the Darkmoon MCP (GPL v3) with no code copied → CeCILL-B license, zero GPL debt. What's reused is the toolbox (sqlmap, nuclei, ffuf, httpx, naabu, katana, whatweb…); the fragile orchestration (opencode + local model) is dropped and replaced by a strong brain that verifies.
Architecture
[Claude Code] --stdio--> [mcp-noctua (host)] --docker.sock--> [darkmoon container = toolbox]
(the brain) (controlled gateway) (sqlmap, nuclei, ffuf…)mcp-noctua reuses the darkmoon container (ascit/darkmoon:latest) as its
toolbox, invoking it via docker.sock. Invoking tools inside a container is not a
derivative work → no license concern. The container is kept alive; noctua starts
it if it's stopped before a run.
Related MCP server: Kali MCP Pentest Server
Exposed MCP tools
Tool | Role |
| Run a whitelisted tool in the toolbox. |
| Bounded crawl (katana). |
| naabu + httpx, bounded. |
| nuclei, bounded. |
| Tools available in the toolbox. |
| Toolbox container state (running / started / not found). |
Guard-rails
Strict allow-list of tools; dangerous patterns blocked (
rm -rf, fork bomb, exfil…).A timeout that actually kills the process inside the toolbox (fixes the Darkmoon flaw).
Authorized testing only; the operator validates every target.
Configuration (.env)
See .env.example. Keys: DOCKER_CONTAINER_NAME, NOCTUA_TIMEOUT,
NOCTUA_REPORTS_DIR, NOCTUA_COMPOSE_DIR, DEBUG.
Install
git clone https://github.com/NeveuGregor/mcp-noctua.git
cd mcp-noctua
python3 -m venv venv && source venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env # adjust as needed
pytestRegister in ~/.claude.json as a stdio MCP server:
venv/bin/python -m src.main (cwd = mcp-noctua).
License
CeCILL-B — see LICENSE.
Available Tools
6 toolshealthA
Etat du conteneur toolbox (present/absent, running, image). Ne demarre rien.
Inputs: no named parameters.
list_toolsA
Liste les outils whitelistes presents (et absents) dans la toolbox.
Inputs: no named parameters.
port_scanB
Scan de ports (naabu top-ports) puis sonde HTTP (httpx) des ports ouverts.
Inputs: target (required), timeout, top_ports.
run_toolA
Execute un outil de reconnaissance/scan whiteliste dans la toolbox (forme exec, sans shell : pas de pipe ni de chainage). Usage test AUTORISE uniquement. Borne par un timeout qui tue reellement le process. Pour enchainer des outils, faire plusieurs appels.
Inputs: command (required), timeout.
vuln_scanB
Scan de vulnerabilites borne (nuclei) : debit/concurrence limites, tags et severite optionnels.
Inputs: url (required), tags, timeout, severity.
web_crawlB
Crawl borne d'une URL (katana, profondeur 1..3, crawling JS).
Inputs: url (required), depth, timeout.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 6 tools
Each tool has a clearly distinct purpose: health checks container state, list_tools enumerates whitelisted tools, port_scan performs port scanning, run_tool executes a generic recon tool, vuln_scan does vulnerability scanning, and web_crawl does web crawling. No overlap or ambiguity.
All names use snake_case and are descriptive, but there's a mix of verb-first (list_tools, run_tool) and noun-first (port_scan, vuln_scan, web_crawl) patterns, plus health stands alone. Still consistent in style and readable.
6 tools is well-scoped for a reconnaissance/scanning toolbox, covering essential functions without being overly numerous or sparse.
The set covers core recon tasks: tool listing, port scanning, vulnerability scanning, web crawling, and a generic run_tool for extensibility. Minor gaps like explicit service detection or DNS enumeration are likely addressable via run_tool.
Maintenance
Related MCP Connectors
Exposes FEDLIN's public security scanners as agent-callable tools over Streamable HTTP.
Find, vet, and run MCP tools through a secure audited gateway with prompt-injection risk scoring
Security scanner for AI agents and MCP servers: injection, jailbreak, tool abuse.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Related MCP Servers
- -licenseNot gradedqualityNot gradedmaintenanceEnables LLMs to execute common security testing tools like nmap, sqlmap, hydra, and metasploit through a secure Docker container. Provides controlled access to penetration testing utilities with configurable timeouts and tool discovery capabilities.-
- FlicenseNot gradedqualityBmaintenanceEnables users to perform security testing and penetration testing tasks through a Docker-based API that provides access to popular security tools like nmap, nikto, sqlmap, wpscan, dirb, and searchsploit. Designed for educational purposes with input sanitization and non-root execution for safer pentesting workflows.-
- FlicenseNot gradedqualityDmaintenanceProvides a secure interface for AI assistants to interact with penetration testing tools like nmap, hydra, sqlmap, and nikto for educational cybersecurity purposes. Includes input sanitization and runs in a Docker container with Kali Linux tools for authorized testing scenarios.-
- AlicenseNot gradedqualityCmaintenanceEnables authorized security testing by letting LLM clients run Kali Linux tools inside an isolated Docker container.MIT