damn-vulnerable-MCP-Server

> An intentionally vulnerable Model Context Protocol server for security training. > Think DVWA but for MCP/AI agent security.

DVMCP is a self-contained training platform for learning how to attack and defend AI agents that use the Model Context Protocol. It simulates a fictional company (NovaTech Solutions) with 6 departments, 28 vulnerable tools, and 38 challenges across 4 difficulty levels.

WARNING: This is intentionally vulnerable software. Do NOT deploy in production. All data is fake.

---

• Prerequisites
• Quick Start
• Local Installation
• Docker
• MCP Inspector
• MCP Host Configuration
• Architecture
• Vulnerability Categories
• Challenges
• Difficulty Levels
• Fake Data
• Project Structure
• Troubleshooting
• Contributing
• License
• Disclaimer

---

• Python 3.11+
• pip (included with Python)
• Docker & Docker Compose (optional, for containerized setup)

---

# Clone the repository
git clone https://github.com/Kyze-Labs/damn-vulnerable-MCP-Server
cd damn-vulnerable-MCP-Server

# Install the core MCP server
pip install -e .

# Run the all-in-one MCP server (all 28 tools)
dvmcp --difficulty beginner

# Or run a specific department
dvmcp --department hr --difficulty beginner

To use the web dashboard, MCP inspector, and exfiltration listener, install with extras:

# Install with all optio