mcp-security-sandbox logo

mcp-security-sandbox

by SirAppSec

An experimental sandbox and lab for exploring MCP hosts, clients, and servers. It allows users to perform attacks against MCP servers and abuse LLMs.

View on GitHub

Last updated: N/A

What is mcp-security-sandbox?

This is an experimental sandbox and lab environment designed to explore and test the security of MCP (Message Channel Protocol) implementations. It provides tools and resources to understand MCP hosts, clients, and servers, and to identify potential vulnerabilities.

How to use mcp-security-sandbox?

To use this sandbox, first install the necessary dependencies using uv install and create a virtual environment with uv venv. Activate the virtual environment and start the MCP server using uv run -- src/mcp-security-sandbox/mcp/github/server.py. Then, run the frontend using streamlit run src/mcp-security-sandbox/frontend/MCP_Chat.py. Ensure that Ollama is installed and its URL is configured in the Ollama client initializations.

Key features of mcp-security-sandbox

  • MCP server implementation (GitHub retrieval)

  • Integration with chat agent playground

  • Burp Suite MCP server interaction

  • Frontend with Streamlit

  • Malicious server creation

Use cases of mcp-security-sandbox

  • Exploring MCP server vulnerabilities

  • Testing MCP client security

  • Developing and testing MCP attack strategies

  • Abusing LLMs through MCP

  • Security research and experimentation

FAQ from mcp-security-sandbox

What is MCP?

MCP stands for Message Channel Protocol. It's a protocol used for communication between different components in a system.

What is Ollama?

Ollama is a tool that allows you to run open-source large language models locally.

How do I install Ollama?

Refer to the Ollama documentation for installation instructions.

How do I configure the Ollama API URL?

The Ollama API URL needs to be set in the Ollama client initializations within the code.

What kind of attacks can be performed in this sandbox?

The sandbox allows for various attacks against MCP servers, including exploiting vulnerabilities and abusing LLMs.