MCP-PST-Server
Enables AI-assisted penetration testing by connecting MCP clients to a Windows API server for executing penetration testing tools like nmap, ffuf, nuclei, and other security tools. Allows AI agents to perform automated security assessments, solve CTF challenges, and assist with ethical hacking tasks through natural language commands.
README
<h1 align="center">MCP-PST-Server</h1>
<p align="center"> <em>Windows AI-Assisted Penetration Testing MCP Server</em> </p>
<p align="center"> <a href="README-ZH.md"><strong>中文文档 </strong></a> | <a href="README.md"><strong>English README</strong></a> </p>
PST MCP Server is a lightweight API bridge designed to connect MCP Clients (e.g., Cherry Studio, Claude Desktop) to an API server, enabling command execution on a Windows terminal. This allows AI Agents or AI Clients to seamlessly invoke various penetration testing tools within the Windows environment, achieving AI-assisted penetration testing.
This enables AI to run terminal tools like nmap, ehole, and others, interact with web applications using tools like curl, gobuster, solve CTF web challenges in real-time, and assist in solving machines from HTB or THM.
Directory Structure
pst_server.py: Windows API Server (Endpoints:/api/tools/<tool>,/api/command,/health)mcp_server.py: MCP Client, encapsulates tools as MCP tools and forwards them to the API Server
🔍 Use Case
The goal is to enable AI-driven offensive security testing by:
- Letting the MCP interact with AI endpoints like OpenAI, Claude, DeepSeek, or any other models.
- Exposing an API to execute commands on a Windows machine.
- Using AI to suggest and run terminal commands to solve CTF challenges or automate penetration testing tasks.
- Allowing MCP apps to send custom requests (e.g.,
curl,nmap,ehole, etc.) and receive structured outputs.
🚀 Features
- 🧠 AI Endpoint Integration: Connect your Windows machine to any MCP of your liking such as Claude Desktop or Cherry-Studio.
- 🖥️ Command Execution API: Exposes a controlled API to execute terminal commands on your Windows machine.
- 🕸️ Web Challenge Support: AI can interact with websites and APIs, capture flags via
curland any other tool AI needs. - 🔐 Designed for Offensive Security Professionals: Ideal for red teamers, bug bounty hunters, or CTF players automating common tasks.
🛠️ Installation
On Pentest-Windows Machine (Will act as PST Server)
git clone https://github.com/arch3rPro/MCP-PST-Server.git
cd MCP-PST-Server
pip install -r requirements.txt
python3 pst_server.py
On your MCP Client (You can run on Windows or Linux)
- You will want to run
python /absolute/path/to/mcp_server.py http://WINDOWS_IP:5100
Configuration for Claude Desktop:
Edit claude_desktop_config.json
{
"mcpServers": {
"pst_mcp": {
"command": "python3",
"args": [
"/absolute/path/to/mcp_server.py",
"--server",
"http://WINDOWS_IP:5100/"
]
}
}
}
Configuration for Cherry Studio:
{
"mcpServers": {
"PST-MCP": {
"name": "pst_mcp",
"type": "stdio",
"isActive": true,
"command": "python3",
"args": [
"/absolute/path/to/mcp_server.py",
"--server",
"http://localhost:5100"
]
}
}
}
Installing Common Penetration Testing Tools (using Scoop and PST-Bucket)
- Install Scoop: https://scoop.sh/
- Add PST-Bucket:
scoop bucket add ar https://github.com/arch3rPro/PST-Bucket
- Install tools (examples, streamline/expand as needed):
scoop install nmap httpx ffuf feroxbuster fscan hydra hackbrowserdatascoop install subfinder dnsx naabu nuclei katanascoop install masscan nikto gobuster john eholescoop install metasploit(if not available, refer to official installer)- Optional:
pip install sqlmap - Netcat: Use
ncat(installed withnmap), ornc(optional)
Recommendation: Use Pentest-Windows Environment Image (with many tools): https://github.com/arch3rPro/Pentest-Windows
Starting the API Server
- Navigate to directory:
/absolute/path/to/MCP-PST-Server - Start:
python pst_server.py --debug --port 5100
- Health Check (PowerShell):
Invoke-RestMethod -Uri http://localhost:5100/health -Method GET
Starting the MCP Client
- Start MCP:
python mcp_server.py --server http://localhost:5100 --debug
Example API Calls (API Server)
-
Nmap:
$body = @{ target="scanme.nmap.org"; scan_type="-sV"; ports=""; additional_args="-T4 -Pn" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/nmap -Method POST -Body $body -ContentType "application/json" -
FFUF:
$body = @{ url="http://target"; wordlist="C:\wordlists\common.txt"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/ffuf -Method POST -Body $body -ContentType "application/json" -
ProjectDiscovery:
# Subfinder $body = @{ domain="example.com"; list_file=""; additional_args="-all -silent" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/subfinder -Method POST -Body $body -ContentType "application/json" # DNSX $body = @{ domain="example.com"; list_file=""; additional_args="-a -resp -silent" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/dnsx -Method POST -Body $body -ContentType "application/json" # Naabu $body = @{ host="example.com"; list_file=""; ports=""; additional_args="-silent" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/naabu -Method POST -Body $body -ContentType "application/json" # Nuclei $body = @{ target="http://example.com"; list_file=""; template=""; tags=""; severity=""; additional_args="-silent" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/nuclei -Method POST -Body $body -ContentType "application/json" # Katana $body = @{ url="http://example.com"; list_file=""; depth="3"; additional_args="-silent" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/katana -Method POST -Body $body -ContentType "application/json" -
Afrog:
$body = @{ target="http://example.com"; list_file=""; pocs=""; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/afrog -Method POST -Body $body -ContentType "application/json" -
Ehole (Fingerprinting):
$body = @{ target="http://example.com"; list_file=""; fingerprints=""; output=""; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/ehole -Method POST -Body $body -ContentType "application/json" -
Nikto:
$body = @{ target="http://example.com"; port=""; ssl=$false; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/nikto -Method POST -Body $body -ContentType "application/json" -
Gobuster (Directory Mode):
$body = @{ mode="dir"; url="http://example.com"; wordlist="C:\wordlists\common.txt"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/gobuster -Method POST -Body $body -ContentType "application/json" -
Masscan:
$body = @{ target="192.168.1.0/24"; ports="80,443"; rate="1000"; iface=""; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/masscan -Method POST -Body $body -ContentType "application/json" -
John (Cracking Example):
$body = @{ hash_file="C:\hashes\passwd.txt"; wordlist="C:\wordlists\rockyou.txt"; format=""; mask=""; rules=$false; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/john -Method POST -Body $body -ContentType "application/json" -
Metasploit (msfconsole -x):
$body = @{ msf_cmd="version; exit"; rc_file=""; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/metasploit -Method POST -Body $body -ContentType "application/json" -
Netcat:
# Listen $body = @{ mode="listen"; listen_port="4444"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/netcat -Method POST -Body $body -ContentType "application/json" # Connect $body = @{ mode="client"; host="127.0.0.1"; port="4444"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/netcat -Method POST -Body $body -ContentType "application/json" -
BBOT (Recursive Internet Scanner):
# Using preset $body = @{ target="example.com"; preset="web-basic"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/bbot -Method POST -Body $body -ContentType "application/json" # Using specific modules $body = @{ target="example.com"; modules="subfinder-enum,dnsx-resolve,nmap-portscan"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/bbot -Method POST -Body $body -ContentType "application/json" # With output directory $body = @{ target="example.com"; preset="cloud-enum"; output_dir="C:\bbot_output"; additional_args="" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/tools/bbot -Method POST -Body $body -ContentType "application/json" -
General Command:
$body = @{ command="whoami" } | ConvertTo-Json Invoke-RestMethod -Uri http://localhost:5100/api/command -Method POST -Body $body -ContentType "application/json"
Security Notes
- Use all tools and features only within legally authorized scopes.
- All command parameters are passed directly to the underlying tools; exercise caution with input to avoid injection and unintended operations.
- Execution timeout defaults to 180s and can be adjusted via
--timeoutat startup.
Iteration Plan
- Add more endpoints based on the tool list in PST-Bucket (e.g.,
afrog,ehole,fscan, etc.). - Enhance parameter validation and output parsing for each tool to improve stability and readability.
⚠️ Disclaimer:
This project is intended solely for educational and ethical testing purposes. Any misuse of the information or tools provided — including unauthorized access, exploitation, or malicious activity — is strictly prohibited. The author assumes no responsibility for misuse.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.