[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tb-a-2077785178964492288-2077785791567757312":3},{"detail":4,"html":18,"toc":19,"catalog":32,"tool":36},{"id":5,"title":6,"coverUrl":7,"url":8,"level":8,"description":9,"publishDate":10,"views":11,"isCollect":12,"isLike":12,"vipRequired":13,"vipFlag":14,"price":8,"contentType":15,"station":8,"seq":15,"readMinutes":16,"toolIds":17,"relatedVideo":8},"2077785791567757312","MCP 构建器：给 AI 造一座桥，接进你的外部系统","",null,"Anthropic 官方出品，装机 9 万+。教你造高质量 MCP 服务器，让 AI 通过设计良好的工具去调外部服务\u002FAPI。支持 Python（FastMCP）和 Node\u002FTypeScript（MCP SDK）双路线，核心标准是能不能让 AI 真正完成实际任务。想把外部系统接进 AI 工作流用它。边界：偏工程，产出是 MCP 服务器代码。","2026-07-17",0,"0",false,"1",1,2,[],"\u003Cdiv style=\"padding:0 4px 24px;font-size:14px;color:#333;line-height:1.85;word-break:break-word\">\n\n\u003Cdiv style=\"background:#FFF7ED;border:1px solid #FFEDD5;border-radius:12px;padding:14px;margin:0 0 18px;font-size:13px;line-height:1.7;color:#9A3412\">🏛️ \u003Cstrong style=\"font-weight:700;color:#EA580C\">官方出品\u003C\u002Fstrong>：Anthropic 官方 skill，装机 \u003Cstrong style=\"font-weight:700;color:#EA580C\">9 万+\u003C\u002Fstrong>。MCP 是 Anthropic 定的协议，造 MCP 服务器认准官方这份最权威。\u003C\u002Fdiv>\n\n\u003Ch1 style=\"font-size:17px;font-weight:800;color:#111;margin:26px 0 14px;padding-left:10px;border-left:5px solid #FF8916;line-height:1.5\" id=\"s1\" data-toc>一、MCP 是什么，为什么要造\u003C\u002Fh1>\n\u003Cp style=\"margin:0 0 14px\">MCP（Model Context Protocol）是让 AI \u003Cstrong style=\"font-weight:700;color:#1a1a1a\">通过设计良好的工具去调外部服务\u002FAPI\u003C\u002Fstrong> 的标准协议。你想让 AI 能操作你的数据库、你的业务系统、某个第三方 API——就给它造一个 MCP 服务器当「桥」。这个 skill 教你怎么造得又对又好用。\u003C\u002Fp>\n\n\u003Ch1 style=\"font-size:17px;font-weight:800;color:#111;margin:26px 0 14px;padding-left:10px;border-left:5px solid #FF8916;line-height:1.5\" id=\"s2\" data-toc>二、两条技术路线都支持\u003C\u002Fh1>\n\u003Cdiv style=\"background:#fff;border:1px solid #F1F5F9;border-radius:14px;padding:14px;margin:0 0 10px;box-shadow:0 2px 10px rgba(0,0,0,0.03)\">\u003Cstrong style=\"font-weight:700;color:#EA580C\">Python 路线\u003C\u002Fstrong>\u003Cdiv style=\"margin-top:4px\">用 FastMCP，快速起一个 MCP 服务器。\u003C\u002Fdiv>\u003C\u002Fdiv>\n\u003Cdiv style=\"background:#fff;border:1px solid #F1F5F9;border-radius:14px;padding:14px;margin:0 0 10px;box-shadow:0 2px 10px rgba(0,0,0,0.03)\">\u003Cstrong style=\"font-weight:700;color:#EA580C\">Node\u002FTypeScript 路线\u003C\u002Fstrong>\u003Cdiv style=\"margin-top:4px\">用官方 MCP SDK。\u003C\u002Fdiv>\u003C\u002Fdiv>\n\n\u003Ch1 style=\"font-size:17px;font-weight:800;color:#111;margin:26px 0 14px;padding-left:10px;border-left:5px solid #FF8916;line-height:1.5\" id=\"s3\" data-toc>三、它强调的核心标准\u003C\u002Fh1>\n\u003Cdiv style=\"background:#F8FAFC;border-left:3px solid #CBD5E1;border-radius:6px;padding:10px 14px;margin:0 0 16px;font-size:13px;line-height:1.7;color:#475569\">衡量一个 MCP 服务器好不好，唯一标准是——\u003Cstrong style=\"font-weight:700;color:#475569\">它能不能让 AI 真正完成实际任务\u003C\u002Fstrong>。所以重点在于工具设计得清不清晰、参数好不好用，而不是堆功能。\u003C\u002Fdiv>\n\n\u003Ch1 style=\"font-size:17px;font-weight:800;color:#111;margin:26px 0 14px;padding-left:10px;border-left:5px solid #FF8916;line-height:1.5\" id=\"s4\" data-toc>四、适合谁 & 边界\u003C\u002Fh1>\n\u003Cp style=\"margin:0 0 14px\">适合\u003Cstrong style=\"font-weight:700;color:#1a1a1a\">想把某个外部系统接进 AI 工作流\u003C\u002Fstrong>的人——比如让 AI 直接读你的业务数据库、调你的内部 API。\u003C\u002Fp>\n\u003Cdiv style=\"background:#FFF7ED;border:1px solid #FFEDD5;border-radius:12px;padding:12px 14px;margin:0 0 16px;font-size:13px;line-height:1.7;color:#9A3412\">⚠️ 三个里最偏工程的一个，产出是 \u003Cstrong style=\"font-weight:700;color:#EA580C\">MCP 服务器代码\u003C\u002Fstrong>，需要一点开发基础。只想用现成能力、不想碰服务端的话，先看「Skill 搜索安装」里有没有现成的集成。\u003C\u002Fdiv>\n\n\u003C\u002Fdiv>\n",[20,23,26,29],{"id":21,"text":22,"level":15},"s1","一、MCP 是什么，为什么要造",{"id":24,"text":25,"level":15},"s2","二、两条技术路线都支持",{"id":27,"text":28,"level":15},"s3","三、它强调的核心标准",{"id":30,"text":31,"level":15},"s4","四、适合谁 & 边界",[33],{"id":5,"title":6,"coverUrl":7,"duration":8,"level":8,"publishDate":10,"views":11,"aiSoftapp":8,"tags":34,"contentType":15,"station":8,"seq":15,"relateId":8,"toolId":35,"readMinutes":16},[],"2077785178964492288",{"id":35,"createTime":37,"createdBy":38,"updateTime":37,"updatedBy":38,"echoMap":39,"slug":40,"name":41,"categorySlug":42,"subCategory":43,"region":8,"website":44,"appName":7,"description":45,"tags":46,"sortOrder":11,"state":15,"tenantCode":52,"tutorialContent":8,"resourceType":53,"accessTypes":7,"installCmd":54,"logo":55,"hotScore":56,"githubStars":11,"category":42},"2026-07-16 23:59:25","1744248479271616512",{},"anthropic-mcp-builder","MCP 构建器","other","综合·元技能","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fskills\u002Ftree\u002Fmain\u002Fskills\u002Fmcp-builder","Anthropic 官方出品，装机 9 万+。教你造高质量的 MCP（Model Context Protocol）服务器，让 AI 能通过设计良好的工具去调外部服务\u002FAPI。支持 Python（FastMCP）和 Node\u002FTypeScript（MCP SDK）两条路线，衡量标准是「能不能让 AI 真正完成实际任务」。想把某个外部系统接进 AI 工作流，用它。边界：偏工程，产出是 MCP 服务器代码。",[47,48,49,50,51],"官方·Anthropic出品","装机9万+","造MCP服务器","接外部API","Python\u002FTS双栈","0000","skill","npx skills add anthropics\u002Fskills@mcp-builder","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4",82]