🔴 RedTeam MCP — Kali Linux Security Assistant & Agentic Server

An LLM-driven Model Context Protocol (MCP) server that connects AI assistants and IDEs (Cursor, Claude Desktop, VS Code, Kiro, Antigravity) to a suite of 42 Kali Linux & bug bounty security tools running inside a container.

It supports both direct tool calls (e.g., "Run an nmap scan on scanme.nmap.org") and autonomous multi-step pentesting powered by local Ollama models (pentest_target).

🛠️ Included Tools & Capabilities (42 MCP Tools)

1. Network Recon & Port Scanning

• nmap_scan: Detailed service and script port scanning.

• rustscan: Ultra-fast port discovery piped to Nmap.

• naabu_scan: Fast SYN-based port scanner, ideal as a first pass before nmap on large ranges.

2. OSINT, DNS & Subdomain Enumeration

• subfinder_enum: Passive subdomain discovery.

• amass_enum: In-depth DNS enumeration & attack surface mapping.

• dnsrecon_enum: DNS record reconnaissance.

• dnsx_resolve: Fast bulk DNS resolution & filtering of subdomain lists.

• alterx_permute: Wordlist-based subdomain permutation for finding non-obvious hosts.

• asnmap_lookup: Maps a domain/org to its ASN and owned IP ranges.

• uncover_search: Queries Shodan/Censys/Fofa-style search engines for exposed hosts.

• theharvester_osint: Public email, IP, and domain harvesting.

• httpx_probe: Live HTTP service probing, status codes, titles, tech-detect.

3. Web Vulnerability & Technology Analysis

• nikto_scan: Web server vulnerability auditing.

• nuclei_scan: Template-based vulnerability scanner (CVEs, exposures, misconfigs).

• whatweb_scan: Web technology & CMS fingerprinting.

• sslscan_audit: SSL/TLS cipher & certificate inspection.

• wpscan_scan: WordPress plugin, theme, and user enumeration.

4. Content Discovery, Crawling & URL Mining

• gobuster_dir: Directory & file brute-forcing.

• ffuf_fuzz: Endpoint, parameter, and header fuzzing.

• feroxbuster: Recursive web content discovery.

• arjun_params: Hidden HTTP parameter discovery.

• katana_crawl: Modern JS-aware crawler for discovering real, linked endpoints.

• gau_urls: Pulls known URLs for a domain from Wayback Machine, Common Crawl & OTX.

• waybackurls_fetch: Fetches archived historical URLs from the Wayback Machine.

5. Vulnerability Exploitation & Injection Testing

• sqlmap_scan: Automated SQL injection detection & testing.

• commix_scan: OS command injection testing.

• dalfox_xss: Automated reflected/DOM XSS scanning and parameter mining.

• hydra_brute: Multi-protocol credential brute-forcing.

• kerbrute: Active Directory Kerberos user enumeration.

• msf_auxiliary: Non-interactive Metasploit module runner.

6. Subdomain Takeover, Secrets & Visual Recon

• subzy_takeover: Detects hijackable/dangling subdomains pointing at unclaimed services.

• gitleaks_scan: Scans for hardcoded secrets, API keys, and credentials.

• gowitness_screenshot: Bulk screenshots live web targets for fast visual triage.

7. Out-of-Band Vulnerability Detection

• interactsh_oob: Blind/OOB interaction detection for SSRF, XXE, and blind RCE via a unique callback domain.

8. Custom Requests

• curl_request: Custom raw HTTP request execution (headers, methods, body, TLS verification toggle).

9. CVE Intelligence & Database

• cve_stats: View local CVE database metrics.

10. Agentic Autonomous Pentesting & Memory

• pentest_target: Plain-English autonomous multi-step security goal planning loop (Ollama).

• generate_report: Autonomous testing + Markdown/HTML report generation.

• export_report: Save penetration testing reports directly to /app/reports/.

• search_findings: Search SQLite history of past scan findings.

• get_attack_surface: Aggregated target intelligence lookup.

• list_sessions: Session history & audit trail.

Related MCP server: Zebbern Kali MCP

💻 Connecting RedTeam MCP to Claude Desktop (Windows)

You can connect RedTeam MCP directly to Claude Desktop on Windows so Claude can use all Kali tools directly in conversation.

Step-by-Step Configuration Guide:

1. Open Claude Desktop Settings:

   • Launch Claude Desktop.

   • Click on your profile/icon or open Settings.

   • Click on Developer in the left menu.

   • Click Edit Config (this opens claude_desktop_config.json in your default text editor).

   (Alternatively, open %APPDATA%\Claude\claude_desktop_config.json in Notepad or File Explorer).

2. Paste the MCP Server Configuration: Add the redteam-kali server entry to your JSON file:

   {
     "mcpServers": {
       "redteam-kali": {
         "command": "docker",
         "args": [
           "exec",
           "-i",
           "redteam-mcp",
           "/app/.venv/bin/python",
           "/app/src/server.py"
         ],
         "disabled": false,
         "autoApprove": []
       }
     }
   }

3. Save & Restart:

   • Save claude_desktop_config.json.

   • Completely close and restart Claude Desktop.

   • You will see a hammer/tool icon indicating redteam-kali tools are connected!

⚙️ Connecting to Other IDEs

Configuration JSON block for all IDEs:

{
  "mcpServers": {
    "redteam-kali": {
      "command": "docker",
      "args": [
        "exec",
        "-i",
        "redteam-mcp",
        "/app/.venv/bin/python",
        "/app/src/server.py"
      ],
      "disabled": false,
      "autoApprove": []
    }
  }
}

🚀 Quick Start Guide

1. Build the Docker Image

docker build -t redteam-mcp:latest .

2. Start the Container

Windows (PowerShell):

.\run.ps1

Linux / macOS:

bash run.sh

3. Verify Container Status

docker ps

The container redteam-mcp will show status Up.

4. Test Server Execution (PowerShell Helper)

.\mcp-helper-fixed.ps1 "nmap scan scanme.nmap.org"

🏗️ Architecture & Modules

redteam-mcp/
├── Dockerfile                  # Kali Linux Docker runtime image
├── requirements.txt            # FastMCP, Ollama SDK, Pydantic, HTTPX, AIOSQLite, Jinja2
├── run.ps1 / run.sh            # Container launch scripts
├── mcp_config.json             # Root IDE MCP configuration template
└── src/
    ├── server.py               # Main MCP Server entry point (42 tools registered)
    ├── agent/                  # PentestAgent Ollama loop & reasoning prompts
    ├── tools/                  # Shell executor, output formatter, tool registry
    ├── parsers/                # Typed Pydantic parsers (Nmap, Nuclei, Nikto, WhatWeb, etc.)
    ├── memory/                 # SQLite store & NVD CVE cross-reference engine
    ├── reporting/              # Markdown & HTML report generators
    └── security/               # Target allowlist, sliding-window rate limiter, JSONL audit logger

🛡️ Security & Target Controls

The server includes built-in safety controls configured via environment variables:

• REDTEAM_ALLOWED_TARGETS: Comma-separated list of IP addresses, CIDR ranges, or domain names permitted for scanning. (Leave empty for open lab testing).

• REDTEAM_RATE_LIMIT: Sliding-window tool call rate limiter per minute (Default: 0 / unlimited).

• REDTEAM_AUDIT_LOG: Path to append-only audit log (/app/data/audit.log).

📋 Changelog note (v2 doc sync)

Compared against the live registered tool list as of this update:

• Added to docs (were live but undocumented): naabu_scan, dnsx_resolve, alterx_permute, asnmap_lookup, uncover_search, katana_crawl, gau_urls, waybackurls_fetch, dalfox_xss, subzy_takeover, gitleaks_scan, gowitness_screenshot, interactsh_oob

• Documented but not currently live — confirm if intentional: wafw00f_detect, find_cves, sync_cve_db

⚠️ Legal Notice

Always ensure you have explicit written authorization before performing security assessments on any host, application, or network. Scanning targets without permission is illegal.