The security scanner codex for AI agents — powered by Scanners-Box.
An MCP (Model Context Protocol) server that turns the Scanners-Box arsenal of 300+ open-source security tools into a queryable knowledge base for Claude, Cursor, and any MCP-compatible AI agent.
Ask your AI assistant "what should I use to scan Kubernetes for misconfigs?" and it will consult the codex, recommend the right tools, show you how to install them, and even run the install for you.
flowchart LR
U([User]) -->|ask in natural language| A[AI Agent\nClaude / Cursor]
A -->|MCP tool call| S[ScanCodex\nMCP Server]
S -->|first run: download & cache| G[(Scanners-Box\nGitHub)]
S -->|query| C[(~/.cache/\nscancodex)]
G --> C
S -->|fetch README| G
S -->|git clone + build| L([Local Machine])
S -->|results| A
A -->|answer| U
| Tool | Description |
|---|---|
list_categories |
Browse all 20 scanner categories |
recommend_scanners |
Find scanners by task description, category, or language |
build_workflow |
Get a full tool chain for a pentest phase |
get_tool_usage |
Fetch install & usage instructions from a tool's GitHub README |
install_tool |
Clone and install a tool locally with one command |
Prerequisites: Python 3.10+
git clone https://github.com/We5ter/ScanCodex
cd ScanCodex
pip install .Scanners-Box data is downloaded automatically on first use and cached to ~/.cache/scancodex/. No extra cloning needed.
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"scancodex": {
"command": "python3",
"args": ["-m", "scancodex.server"]
}
}
}claude mcp add scancodex -- python3 -m scancodex.serverWhat tools should I use to test an LLM app for prompt injection?
I need to scan a Kubernetes cluster for security issues — what do you recommend?
Build me a recon workflow for a pentest engagement.
Show me Go-based vulnerability scanners for container images.
How do I install and use GitHack?
Install subfinder for me.
recon · vuln_scan · web · container · mobile · smart_contract · ai_apps · malware · code_analysis · incident · a3c
MIT — data sourced from Scanners-Box (CC-BY-NC-ND-4.0).
