📄 中文摘要
在构建 AI 集成时，常常会遇到敏感数据的问题，例如客户姓名、预订参考、病人记录等。直接将这些数据传输到 OpenAI 或 Anthropic 的推理端点并不安全，尽管常规建议是避免包含敏感数据，但实际工作流程中上下文是 LLM 有效性的关键。为了解决这一问题，可以采取先清洗数据再发送的策略。该教程将指导用户如何在 15 分钟内为任何 AI 集成添加隐私保护层。
如何使用 MCP 服务器和 LLM 代理构建隐私安全的 AI 集成
📄 English Summary
How to Build Privacy-Safe AI Integrations with MCP Servers and LLM Agents
Building AI integrations often encounters the challenge of handling sensitive data, such as guest names, booking references, patient records, and more. Directly transmitting this data to OpenAI or Anthropic's inference endpoints poses privacy risks, and while the standard advice is to avoid including sensitive information, real workflows rely heavily on context for LLM effectiveness. A more effective solution is to scrub the data before sending it. This tutorial provides guidance on adding a privacy layer to any AI integration in just 15 minutes.