我开发 MCP Guard 是因为让 AI 代理直接访问数据库让我感到害怕。
1 分•作者: pritahir•6 个月前
模型上下文协议功能强大，但未经防护地将你的生产数据库连接到 AI 代理是令人恐惧的。<p>我构建了 MCP Guard (https://mcp-shield.vercel.app)，因为我希望能够了解我的 AI 代理实际在做什么，并能够控制它们，在执行危险命令之前将其阻止。<p>它是一个托管的 SaaS 仪表板——无需 NPM 包，无需维护本地代理。只需通过浏览器配置安全规则，然后将你的 AI 客户端指向安全端点即可。
查看原文
The Model Context Protocol is powerful, but connecting your production database to an AI agent without guardrails is terrifying.<p>I built MCP Guard (https://mcp-shield.vercel.app) because I wanted visibility into what my AI agents were actually doing, and control to block dangerous commands before they execute.<p>It's a hosted SaaS dashboard - no NPM packages, no local proxies to maintain. Just configure security rules through a browser and point your AI client to the secure endpoint.