The Model Context Protocol (MCP) is quickly becoming essential for connecting large language models (LLMs) to various tools and data sources. As its adoption rises, so does the need for robust security measures. This cheat sheet outlines specific steps security teams can take to safeguard MCP implementations while maintaining the pace of innovation. Key recommendations include securing MCP servers and supply chains to prevent unauthorized access and data breaches. It emphasizes the importance of enforcing least-privilege access for tokens and tools, ensuring that users and applications have only the permissions necessary to perform their functions. This minimizes the risk of misuse or accidental exposure of sensitive information. The cheat sheet also highlights the necessity of human-in-the-loop safeguards. These measures help mitigate potential errors that can arise from automation, providing an additional layer of oversight. By implementing these practices, organizations can better protect their MCP systems and the valuable data they handle, all without stifling the benefits that come from leveraging LLMs and connected tools.