Smart Search MCP

Instructions

Implementation Reference

• mcp_server_node.js:525-597 (handler)

  Handler for 'ai_search_npm' tool: validates input, generates NPM web and registry search URLs, creates detailed markdown with WebFetch prompts and search tips, saves to file, returns response.

  case 'ai_search_npm': {
    const rawQuery = normalizeString(args.query);
    const resolvedSize = clampNumber(args.size, 1, 100, 10);
  
    if (!rawQuery) {
      throw new Error('搜索关键词不能为空');
    }
  
    const searchUrl = `https://www.npmjs.com/search?q=${encodeURIComponent(rawQuery)}`;
    const registryUrl = `https://registry.npmjs.org/-/v1/search?text=${encodeURIComponent(rawQuery)}&size=${resolvedSize}`;
  
    // NPM 搜索技巧
    const tips = [
      `精确包名: ${rawQuery} (使用完整包名)`,
      `关键词搜索: keywords:${rawQuery}`,
      `作者搜索: author:${rawQuery}`,
      `维护者: maintainer:${rawQuery}`,
      `作用域包: @scope/${rawQuery}`,
      `特定版本: ${rawQuery}@latest`
    ];
  
    // 相关搜索建议
    const relatedSearches = [
      `${rawQuery} typescript`,
      `${rawQuery} cli`,
      `${rawQuery} plugin`,
      `@types/${rawQuery}`
    ];
  
    // 热门类别推荐
    const categories = [
      'react', 'vue', 'express', 'webpack',
      'babel', 'eslint', 'testing', 'cli-tools'
    ];
  
    const detailsContent = `📦 NPM 包搜索\n\n` +
      `**搜索关键词**: ${rawQuery}\n` +
      `**期望结果数**: ${resolvedSize} 个\n\n` +
      `---\n\n` +
      `🔗 **网页搜索**: ${searchUrl}\n` +
      `🔗 **API搜索**: ${registryUrl}\n\n` +
      `⚠️ **请使用 WebFetch 工具获取搜索结果**:\n` +
      `\`\`\`javascript\n` +
      `// 方式1: 网页搜索\n` +
      `WebFetch({\n` +
      `  url: "${searchUrl}",\n` +
      `  prompt: "提取前${resolvedSize}个包的：包名、描述、版本号、周下载量、最后更新时间"\n` +
      `})\n\n` +
      `// 方式2: API搜索 (推荐，结构化数据)\n` +
      `WebFetch({\n` +
      `  url: "${registryUrl}",\n` +
      `  prompt: "解析JSON数据，提取包的名称、描述、版本、作者和下载统计"\n` +
      `})\n` +
      `\`\`\`\n\n` +
      `---\n\n` +
      `💡 **NPM 搜索技巧**:\n` +
      tips.map(tip => `• ${tip}`).join('\n') +
      `\n\n📌 **相关搜索建议**:\n` +
      relatedSearches.map(s => `• ${s}`).join('\n') +
      `\n\n🏷️ **热门包分类**:\n` +
      categories.map(cat => `• ${cat}`).join(' ') +
      `\n\n📚 **直接访问包详情**: https://www.npmjs.com/package/${rawQuery}`;
  
    const filepath = await saveSearchResult('npm-search', rawQuery, detailsContent);
  
    return makeTextResponse(
      `📦 **NPM包搜索**\n\n` +
      `**关键词**: ${rawQuery}\n` +
      `**搜索链接**: ${searchUrl}\n\n` +
      `✅ 详细信息已保存至: ${filepath || '保存失败'}\n` +
      `💡 使用 WebFetch 工具访问搜索链接获取结果`
    );
  }

• mcp_server_node.js:117-128 (registration)

  Registration of 'ai_search_npm' tool in AI_TOOLS array: includes name, description, and input schema definition.

  {
    name: 'ai_search_npm',
    description: '📦 NPM包搜索 - 搜索NPM包和相关文档\n\n【重要】此工具会返回NPM搜索URL，Claude Code应该使用WebFetch工具访问该URL以获取真实搜索结果。',
    inputSchema: {
      type: 'object',
      properties: {
        query: { type: 'string', description: '包名或关键词' },
        size: { type: 'number', description: '返回结果数量，默认10', default: 10 }
      },
      required: ['query']
    }
  },

• mcp_server_node.js:120-128 (schema)

  Input schema for 'ai_search_npm' tool defining query (required string) and optional size (number).

    inputSchema: {
      type: 'object',
      properties: {
        query: { type: 'string', description: '包名或关键词' },
        size: { type: 'number', description: '返回结果数量，默认10', default: 10 }
      },
      required: ['query']
    }
  },