damn-vulnerable-MCP-server

DVMCP provides a deliberately vulnerable implementation of the Model Context Protocol (MCP) to help security researchers, developers, and AI safety professionals learn about potential security issues. It features 10 challenges of increasing difficulty, showcasing various types of vulnerabilities and attack vectors that can occur in MCP implementations. Users can explore these vulnerabilities in a controlled environment, gaining insights into how to mitigate them effectively. The project serves as a practical resource for understanding the implications of using MCP in applications that interact with Large Language Models (LLMs). By engaging with the challenges presented, users can develop a deeper understanding of security considerations, including prompt injection, tool poisoning, and excessive permissions, ultimately enhancing their skills in securing LLM applications.

{
  "mcpServers": {
    "damn-vulnerable-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-damn-vulnerable-mcp-server"
      ]
    }
  }
}