cybersec-toolkit — Claude Code MCP server: features and setup
This MCP server lets an agent find and run security tools for CTFs, pentests, and DFIR. External targets and script execution need explicit permission.
claude mcp add --env CYBERSEC_MCP_ALLOW_EXTERNAL=<CYBERSEC_MCP_ALLOW_EXTERNAL> --env CYBERSEC_MCP_ALLOW_SCRIPTS=<CYBERSEC_MCP_ALLOW_SCRIPTS> --transport stdio cybersec-toolkit -- docker run -i --rm -e CYBERSEC_MCP_ALLOW_EXTERNAL -e CYBERSEC_MCP_ALLOW_SCRIPTS ghcr.io/26zl/cybersec-toolkit:1.4.0What cybersec-toolkit does
cybersec-toolkit exposes 15 MCP tools for tool discovery, recommendations, and governed execution. Its collection covers 670+ security tools for CTF, pentest, DFIR, and blue-team work. An agent can select tools for a problem and work through it step by step.
The server runs in a disposable Kata Containers VM by default. Sandbox mode requires Linux with KVM, Docker 23+ with a Kata runtime, and Node.js 22+; use host mode on macOS or Windows. External targets and script execution are disabled by default. Those capabilities require explicit authorization.
What it helps with
- find tools for a CTF challenge
- investigate a suspicious binary
- map a lab network’s attack surface
- select tools for a DFIR investigation
Example prompts for the agent
- Analyze this binary and suggest suitable tools for the investigation.
- Map the attack surface of my lab network at 10.10.0.0/24.
- Find tools for this incident investigation and propose a sequence of steps.
How to install cybersec-toolkit
Platform CLI
claude mcp add --env CYBERSEC_MCP_ALLOW_EXTERNAL=<CYBERSEC_MCP_ALLOW_EXTERNAL> --env CYBERSEC_MCP_ALLOW_SCRIPTS=<CYBERSEC_MCP_ALLOW_SCRIPTS> --transport stdio cybersec-toolkit -- docker run -i --rm -e CYBERSEC_MCP_ALLOW_EXTERNAL -e CYBERSEC_MCP_ALLOW_SCRIPTS ghcr.io/26zl/cybersec-toolkit:1.4.0Replace <…> with your own values
Platform CLI
codex mcp add cybersec-toolkit --env CYBERSEC_MCP_ALLOW_EXTERNAL=<CYBERSEC_MCP_ALLOW_EXTERNAL> --env CYBERSEC_MCP_ALLOW_SCRIPTS=<CYBERSEC_MCP_ALLOW_SCRIPTS> -- docker run -i --rm -e CYBERSEC_MCP_ALLOW_EXTERNAL -e CYBERSEC_MCP_ALLOW_SCRIPTS ghcr.io/26zl/cybersec-toolkit:1.4.0Replace <…> with your own values
Config file
{
"mcpServers": {
"cybersec-toolkit": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"CYBERSEC_MCP_ALLOW_EXTERNAL",
"-e",
"CYBERSEC_MCP_ALLOW_SCRIPTS",
"ghcr.io/26zl/cybersec-toolkit:1.4.0"
],
"env": {
"CYBERSEC_MCP_ALLOW_EXTERNAL": "<CYBERSEC_MCP_ALLOW_EXTERNAL>",
"CYBERSEC_MCP_ALLOW_SCRIPTS": "<CYBERSEC_MCP_ALLOW_SCRIPTS>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"cybersec-toolkit": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"CYBERSEC_MCP_ALLOW_EXTERNAL",
"-e",
"CYBERSEC_MCP_ALLOW_SCRIPTS",
"ghcr.io/26zl/cybersec-toolkit:1.4.0"
],
"env": {
"CYBERSEC_MCP_ALLOW_EXTERNAL": "<CYBERSEC_MCP_ALLOW_EXTERNAL>",
"CYBERSEC_MCP_ALLOW_SCRIPTS": "<CYBERSEC_MCP_ALLOW_SCRIPTS>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e CYBERSEC_MCP_ALLOW_EXTERNAL=<CYBERSEC_MCP_ALLOW_EXTERNAL> -e CYBERSEC_MCP_ALLOW_SCRIPTS=<CYBERSEC_MCP_ALLOW_SCRIPTS> cybersec-toolkit docker run -i --rm -e CYBERSEC_MCP_ALLOW_EXTERNAL -e CYBERSEC_MCP_ALLOW_SCRIPTS ghcr.io/26zl/cybersec-toolkit:1.4.0Replace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about cybersec-toolkit
How do I install cybersec-toolkit in Claude Code?
Run `claude mcp add --env CYBERSEC_MCP_ALLOW_EXTERNAL=<CYBERSEC_MCP_ALLOW_EXTERNAL> --env CYBERSEC_MCP_ALLOW_SCRIPTS=<CYBERSEC_MCP_ALLOW_SCRIPTS> --transport stdio cybersec-toolkit -- docker run -i --rm -e CYBERSEC_MCP_ALLOW_EXTERNAL -e CYBERSEC_MCP_ALLOW_SCRIPTS ghcr.io/26zl/cybersec-toolkit:1.4.0`. Restart Claude Code to load the MCP tools.
What does cybersec-toolkit do?
It gives an AI client tools to discover and run security utilities for work such as CTFs, pentests, and digital forensics. Execution is governed, while external targets and arbitrary scripts are disabled unless authorized.
What does sandbox mode require?
It requires Linux with KVM, Docker 23+ with a Kata runtime, and Node.js 22+. On macOS, Windows, or virtual machines without nested virtualization, use host mode.
