Installation and Flags

Requirements

• Python 3.13+ (recommended)
• Linux (Kali, Ubuntu, or Debian recommended)
• External security tools installed via nyxstrike.sh -t or manually

---

Scripted Setup (recommended)

git clone https://github.com/CommonHuman-Lab/nyxstrike.git
cd nyxstrike

./nyxstrike.sh -a                  # Setup + start server
./nyxstrike.sh -a -t               # + install external tools
./nyxstrike.sh -a -t -ai           # + local AI model (~8.4 GB RAM)
./nyxstrike.sh -a -t -ai-small     # + smaller AI model (~2.5 GB RAM)

# Docker
docker compose up --build -d    # Build + start
docker compose down             # Stop

---

Flag Reference

Flag	Long form	Description
-a	--all	Auto-setup: create venv, install Python deps, start server
-t	--install-tools	Install external security tools (apt, cargo, git clones)
-b	--install-big-packages	Install heavy Python extras — implies -t
-s	--update-self	Run git pull on the NyxStrike repo itself
-y	--update-python-packages	Force reinstall all Python requirements
-p <bin>	--python <bin>	Override the Python binary used (e.g. python3.11)
-ai		Install Ollama and pull huihui_ai/gemma-4-abliterated:e2b (~8.4 GB RAM required)
-ai-small		Install Ollama and pull huihui_ai/qwen3.5-abliterated:4b (~2.5 GB RAM)
--server		Start the NyxStrike API server only
--mcp		Start the MCP client only
--server-url <url>		Override server URL for MCP (default: http://127.0.0.1:8888)
--profile <name>		MCP tool profile to load (default: default)
-h	--help	Print usage

Default behavior (no flags): Starts the MCP client only — useful for 5ire and similar clients that manage the process themselves.

---

Common Combinations

# Full setup from scratch, install tools, start server
./nyxstrike.sh -a -t

# Full setup with local AI model (requires ~8.4 GB RAM)
./nyxstrike.sh -a -t -ai

# Full setup with smaller local AI model (~2.5 GB RAM)
./nyxstrike.sh -a -t -ai-small

# Update the repo and restart the server
./nyxstrike.sh -s --server

# Reinstall Python packages after a dependency change
./nyxstrike.sh -y --server

---

MCP Profiles

The --profile flag controls which tool categories are loaded into the MCP client. Smaller profiles reduce context size for lightweight or local LLMs.

Profile	Contents
default	Core tools: network scan, recon, web, vuln scan, exploitation, password cracking, WiFi, monitoring
full	Every registered tool category
compact	Gateway tools only (classify_task + run_tool) — for small/local LLMs
net_scan	nmap, rustscan, masscan, arp-scan
recon	amass, subfinder, assetfinder, shuffledns, massdns, autorecon, theHarvester
web_fuzz	dirb, ffuf, dirsearch, gobuster, feroxbuster, dotdotpwn, wfuzz
web_scan	nikto, sqlmap, wpscan, jaeles, dalfox, burpsuite, zap, xsser
wifi_pentest	Full aircrack-ng suite + hcxdumptool, eaphammer, wifite2, bettercap, mdk4
password_cracking	hydra, john, hashcat, medusa, patator, hashid, ophcrack
binary_analysis	binwalk, checksec, strings, objdump, ghidra, angr, autopsy, libc-database, xxd
cloud_audit	prowler, scout-suite
osint	sherlock, spiderfoot, sublist3r, parsero, joomscan

See all profiles in mcp_core/tool_profiles.py.

---

MCP compact mode

Activate Minimal Mode for the MCP client:

• Loads only the two essential gateway tools: classify_task and run_tool.
• Ideal for lightweight deployments, automation pipelines, or resource-constrained systems.
• Streamlined integration—perfect for embedding NyxStrike into other platforms or CI/CD workflows.

Example usage

nyxstrike-env/bin/python3 nyxstrike_mcp.py --compact

Example usage

  "servers": {
    "nyxstrike": {
      "type": "stdio",
      "command": "/path/to/nyxstrike/nyxstrike-env/bin/python3",
      "args": [
        "/path/to/nyxstrike/nyxstrike_mcp.py",
        "--server",
        "http://127.0.0.1:8888",
        "--compact"
      ]
    }
  },
  "inputs": []
} 

---

Environment Variables

Variable	Default	Purpose
NYXSTRIKE_PORT	8888	API server port
NYXSTRIKE_HOST	127.0.0.1	Bind address (0.0.0.0 for network access)
NYXSTRIKE_API_TOKEN	(none)	Bearer auth token — enables auth on all endpoints when set
NYXSTRIKE_LLM_PROVIDER	ollama	LLM backend: ollama, openai, anthropic
NYXSTRIKE_LLM_MODEL	huihui_ai/qwen3.5-abliterated:9b	Model name
NYXSTRIKE_LLM_URL	http://localhost:11434	Ollama base URL
NYXSTRIKE_LLM_API_KEY	(none)	Required for OpenAI or Anthropic
NYXSTRIKE_DATA_DIR	.nyxstrike_data/	Persistent data directory (SQLite DB, run history)
COMMAND_TIMEOUT	500	Default command timeout in seconds
DEBUG_MODE	0	Enable Flask debug mode

Manual Setup

# 1) Clone
git clone https://github.com/CommonHuman-Lab/nyxstrike.git
cd nyxstrike

# 2) Create and activate venv
python3 -m venv nyxstrike-env
source nyxstrike-env/bin/activate

# 3) Install dependencies
pip3 install -r requirements.txt

# 4) Start API server + dashboard (http://localhost:8888)
python3 nyxstrike_server.py

# 5) In a second terminal, start MCP client
nyxstrike-env/bin/python3 nyxstrike_mcp.py --server http://localhost:8888 --profile full