📄 中文摘要
MCP 服务器已迅速成为现代 AI 系统的重要构建块。MCP（模型上下文协议）允许模型以结构化方式与外部工具和服务进行交互，超越了静态提示的限制。通过连接工具并暴露其架构，模型可以几乎立即开始使用这些工具。然而，当 MCP 工具进入生产环境时，架构的重要性显现出来，随之而来的问题包括谁控制代理可以使用哪些工具、如何确保安全性、如何进行扩展以及如何实现治理等。这些问题在生产环境中尤为关键，必须认真考虑以确保系统的有效性和安全性。
如何在生产环境中运行 MCP 服务器（AI 工具的安全性、扩展性与治理）
📄 English Summary
How to Run MCP Servers in Production (Security, Scaling & Governance for AI Tooling)
MCP servers have quickly become a crucial building block in modern AI systems. The Model Context Protocol (MCP) allows models to interact with external tools and services in a structured manner, moving beyond static prompts. By connecting tools and exposing their schemas, models can start utilizing them almost immediately. However, once MCP tooling enters production environments, the architecture becomes significant. Questions arise regarding who controls which tools an agent can use, how to ensure security, how to scale effectively, and how to implement governance. These issues are particularly critical in production settings and must be carefully addressed to ensure the effectiveness and security of the system.