📄 中文摘要
构建出色的 MCP 服务器后，面临的一个主要问题是其难以被发现。尽管 MCP 解决了 AI 模型与工具之间的接口问题，但并未解决发现问题。代理 A 如何找到代理 B？如何调用在他人基础设施上运行的 MCP 服务器？目前的解决方案是无法实现的，用户需要知道确切的 URL，手动共享配置，并寄希望于不发生变化。这种情况类似于 DNS 出现之前的网络状态，缺乏有效的名称解析层使得资源的发现变得困难。
你的 MCP 服务器没有名字：代理身份的重要性
出处: Your MCP Server Has No Name: Why Agent Identity Matters
发布: 2026年2月15日
📄 English Summary
Your MCP Server Has No Name: Why Agent Identity Matters
After building a great MCP server, a significant issue arises: it is hard to find. While MCP addresses the interface problem for AI models to interact with tools, it does not resolve the discovery problem. How can Agent A find Agent B? How do you call an MCP server running on someone else's infrastructure? Currently, the answer is that you can't; users need to know the exact URL, share configurations manually, and hope nothing changes. This situation mirrors the state of the web before the advent of DNS, where the lack of an effective naming layer made resource discovery challenging.