[{"_1":2,"_328":-5,"_329":-5},"loaderData",{"_3":4,"_20":21,"_22":23,"_24":25,"_259":260},"root",{"_5":6,"_7":8},"toasts",[],"visitor",{"_9":10,"_11":12},"viewportWidth",1280,"visitorSession",{"_13":14,"_16":-5,"_17":18,"_19":-5},"attributes",[15],"BOT","membership","uid","zyM5aHVzIV","userAccount","routes/_public/_layout",{},"routes/_public/mcp/servers/~namespace/_layout",{},"routes/_public/mcp/servers/~namespace/~slug/_pages/_layout",{"_26":-5,"_27":28,"_29":30,"_178":179},"githubUser","maintainer",false,"mcpServer",{"_31":32,"_33":34,"_13":60,"_65":66,"_67":66,"_68":69,"_74":75,"_98":-5,"_99":100,"_101":28,"_102":103,"_84":114,"_115":89,"_116":-5,"_117":-5,"_118":119,"_120":121,"_128":-5,"_129":130,"_135":-5,"_110":85,"_136":137,"_140":141,"_17":175,"_176":177},"addedAt",["D",1739336405764],"argumentsJsonSchema",{"_35":36,"_57":58,"_41":59},"properties",{"_37":38,"_43":44,"_46":47,"_49":50,"_52":53},"METASO_SID",{"_39":40,"_41":42},"description","SID for Metaso search","type","string","METASO_UID",{"_39":45,"_41":42},"UID for Metaso search","BOCHA_API_KEY",{"_39":48,"_41":42},"API key for Bocha search","BRAVE_API_KEY",{"_39":51,"_41":42},"API key for Brave Search","SEARCH_ENGINE",{"_54":55,"_39":56,"_41":42},"default","bocha","The search engine to use","required",[],"object",[61,62,63,64],"category:search","hosting:hybrid","language:python","license_grade:a","descriptionMarkdown","The Search MCP Server enables seamless integration of network and local search capabilities in tools like Claude Desktop and Cursor, utilizing the Brave Search API for high-concurrency and asynchronous requests.","descriptionPlainText","githubProject",{"_70":71,"_72":73},"readme","# Search MCP Server\n\n![image](./images/logo.png)\n\n一个基于MCP协议的搜索服务实现，提供多种搜索引擎支持，Cursor和Claude Desktop能与之无缝集成。\n\n使用Python开发，支持异步处理和高并发请求，目前支持三种搜索引擎选择：\n\n- Brave Search ：国外一家专业提供搜索接口服务产品\n\n- 秘塔（Metaso）搜索：秘塔AI搜索的逆向实现接口，非官方接口\n\n- 博查（bocha）搜索：国内Search API市场占有率最高的搜索API产品\n\n更多MCP知识，见AI全书([ 一文看懂什么是MCP(大模型上下文)？用来干什么的？怎么用它？](https://aibook.ren/archives/mcp-course))\n\n**作者**：凌封（微信fengin）\n\n**网址**：[https://aibook.ren]()（AI全书）\n\n## 使用示例\n\n![image](./images/demo.png)\n\n## 功能特点\n\n- **多搜索引擎支持**: \n - Brave Search: 提供网络搜索和位置搜索\n - Metaso搜索: 提供网络搜索和学术搜索，支持简洁和深入两种模式\n - 博查搜索: 提供网络搜索，支持时间范围过滤、详细摘要和图片搜索\n- **适用场景**: Claude Desktop或者Cursor无缝集成使用，大大扩展工具的内容获取能力\n- **模块化设计**: 每个搜索引擎都是独立的模块，也可以单独拷出去其他地方使用\n\n## 三种搜索的选择\n\n运行时只能生效一种搜索引擎，为了方便大家选择配置哪个上线，我列了下大致的对比：\n\n| 搜索引擎 | 国内/外 | 需魔法 | 自带总结 | 质量 | 免费 | 官方 | 速度 | 注册门槛 |\n| ------ | ---- | --- | ---- | --- | ----- | --- | ------- | ---- |\n| Brave | 国外 | 是 | 否 | 高 | 是(限量) | 是 | 中 | 很高 |\n| Metaso | 国内 | 否 | 是 | 中 | 是 | 否 | 慢(AI总结) | 低 |\n| Bocha | 国内 | 否 | 否 | 高 | 否 | 是 | 极快 | 低 |\n\n## 安装和使用\n\n### 1. 环境要求\n\n- Python 3.10+\n- uv 0.24.0+\n- node.js v20.15.0\n- cursor >=0.45.10 (低于该版本mcp server配置老是连不上)\n- 科学上网（仅使用Brave Search需要）\n\n#### 1.1 安装浏览器驱动(仅Metaso需要)\n\n```\n# 安装Playwright框架\npip install playwright>=1.35.0\n# 安装浏览器驱动,仅安装chromium\nplaywright install chromium\n```\n\n\n\n### 2.下载代码\n\n```bash\ngit clone https://github.com/fengin/search-server.git\n```\n\n### 3. 启用你要的搜索引擎\n\n打开项目根目录，修改server.py以下代码选择启用类型：\n\n```python\n# 搜索引擎配置\nSEARCH_ENGINE = os.getenv(\"SEARCH_ENGINE\", \"bocha\")\n```\n\n其中值分别对应有brave、metaso、bocha，也可以通过配置环境变量SEARCH_ENGIN\n\n### 4\\. 配置对应的搜索模块\n\n以下三个模块目录下都对应有一个config.py文件：\n\n- src\\search\\proxy\\brave\n\n- src\\search\\proxy\\metaso\n\n- src\\search\\proxy\\bocha\n\n根据你的选择，修改对应的config.py文件配置\n\n#### 4.1 brave search配置\n\n```python\n# 检查API密钥\nBRAVE_API_KEY = os.getenv(\"BRAVE_API_KEY\")\nif not BRAVE_API_KEY:\n BRAVE_API_KEY = \"你申请的 brave_api_key\"\n```\n\n在Claude Desktop里面使用的话，也可以在Claude Desktop里配置通过环境变量传这个参数，但是Cursor目前不支持环境变量，只能在这文件里修改\n\nAPI KEY 申请地址：[Brave Search - API](https://api-dashboard.search.brave.com/login)\n\n申请门槛比较高，要求：\n\n- 魔法（使用时也需要）\n\n- 邮箱验证\n\n- 信用卡（可以用虚拟的：[https://cardgenerator.org/](https://cardgenerator.org/)）\n\n#### 4.2 秘塔(metaso)配置\n\n```python\n# 认证信息\nMETASO_UID = os.getenv(\"METASO_UID\")\nMETASO_SID = os.getenv(\"METASO_SID\")\nif not METASO_UID or not METASO_SID:\n METASO_UID = \"你获取的 metaso_uid\"\n METASO_SID = \"你获取的 metaso_sid\"\n```\n\n同样Claude Desktop使用可以通过 MCP Servers配置里环境变量；\n\n**uid和sid获取方式：**\n\n进入秘塔AI搜索，登录账号（建议登录账号，否则可能遭遇奇怪的限制），然后F12打开开发者工具，从Application > Cookies中找到`uid`和`sid`的值。\n\n![获取uid-sid](./images/meta-uid-sid.png)\n\n**多账号接入**\n\n注意：目前怀疑秘塔对IP地址的总搜索次数有限制，建议加入IP轮换\n\n你可以通过提供多个账号的uid-sid并使用`,`修改下相关的使用代码，每次请求服务会从中挑选一个，本人后续再考虑。\n\n#### 4.3 博查(bocha)配置\n\n```python\nBOCHA_API_KEY = os.getenv(\"BOCHA_API_KEY\", \"\")\nif not BOCHA_API_KEY:\n BOCHA_API_KEY=\"你申请的 bocha_api_key\"\n```\n\n注册申请地址：https://open.bochaai.com/\n\n调用按次数收费，不便宜，但是搜索质量确实比较好，我这有少量的免费试用码，有需要的微信联系我；\n\n### 5. AI工具配置\n\n#### 5.1 Cursor里配置\n\n![Cursor配置](./images/cursor-mcp-server.png)\n\n- name: search\n\n- type: cmd\n\n- command: uv --directory D:\\\\code\\\\search-server run search\n\n其中 “D:\\code\\search-server” 就是你拉下来源代码目录\n\n#### 5.2 Claude Desktop配置\n\n找到配置文件\n\n**方法一**\n\n```\n# widnows\nC:\\Users\\{用户}\\AppData\\Roaming\\Claude\\claude_desktop_config.json\n# mac/linux 应该在用户家目录下找\n```\n\n**方法二**\n\n打开Claude Desktop应用进入查看：\nClaude Desktop—>菜单—>Settings—>Developer—>Edit Config\n\n编辑增加以下MCP Server:\n\n```json\n{\n \"mcpServers\": {\n \"search\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"D:\\\\code\\\\search-server\",\n \"run\",\n \"search\"\n ],\n \"env\": {\n \"BRAVE_API_KEY\": \"你申请的API KEY\"\n }\n }\n }\n}\n```\n\n环境变量看你需要，如果代码改了这就没必要配置了\n\nCursor会弹出一个黑窗口，不要关闭，不要关闭，这是启动的MCP Server进程，目前没办法解决不弹出来。\n\nClaude Desktop配置完一定要重启应用才生效。\n\n#### 5.4 问题排错\n\n配置Cursor后很多人遇到，MCP Servers里面配置完，还是显示状态红点，Tools Not Found，使用时也不会去调用，这是因为根本没配置好。\n\n大部分可能出现的情况是：\n\n1. 环境没准备好，包括需要的软件及版本要求，详见环境章节\n\n2. 准备的环境不对，像windows有cmd终端、powershell终端、还有可能装了gitbash终端，你打开cmd终端（cursor一般是这个）检查下环境，直接运行下uv --directory D:\\code\\search-server run search\n\n3. 配置路径/命令不对，可以打开终端运行命令看下：uv --directory D:\\code\\search-server run search\n\n4. 把黑窗口关闭了，再次开启需要重启Cursor\n\n5. Cursor版本太旧\n\n6. 运行时报以下错，原因是没有安装chromium，解决办法见环境准备1.1章节\n \n ```shell\n 错误:搜索执行错误:BrowserType.launch persistent context:Executable doesn't exist atC:\\Users\\fengi\\AppDatalLocal\\ms-playwright\\chromium headless shell-1155\\chrome-winlheadless shell.exe\n ```\n\n### 6. 使用\n\n在你的Claude Desktop 或者 Cursor 里直接你正常的工作就行了，必要时它会自动调用搜索接口获取内容，比如说，你整理下网络上2025年技术发展方向作为软件什么内容，它就会去调用搜索工具获取网络信息：\n\n- 配置好工具后，它的信息里面就知道有这个工具\n\n- 根据你的要求，它自动会分析判断需要用到搜索工具\n\n- 根据需求，提取关键词，调用搜索工具\n\n- 根据搜索返回内容，组织你要的结果\n\n有一点需要注意的，在Cursor里面，必须启用composer的agent模式工作才会生效，调用工具时，也需要你点下执行；\n\n## 技术内幕\n\n### 项目结构\n\n```shell\nsearch/\n├── __init__.py\n├── server.py # MCP服务器实现\n└── proxy/ # 搜索引擎代理\n ├── brave/ # Brave搜索模块\n │ ├── __init__.py\n │ ├── client.py # 核心客户端实现\n │ ├── config.py # 配置和速率限制\n │ └── exceptions.py # 异常定义\n ├── metaso/ # Metaso搜索模块\n │ ├── __init__.py\n │ ├── client.py # 核心客户端实现\n │ ├── config.py # 配置和速率限制\n │ └── exceptions.py # 异常定义\n ├── bocha/ # 博查搜索模块\n │ ├── __init__.py\n │ ├── client.py # 核心客户端实现\n │ ├── config.py # 配置和速率限制\n │ └── exceptions.py # 异常定义\n ├── brave_search.py # Brave MCP工具实现\n ├── metaso_search.py # Metaso MCP工具实现\n └── bocha_search.py # 博查搜索MCP工具实现\n```\n\n### 接口参数\n\n#### Brave Search引擎\n\n- **search**\n \n - 执行网络搜索，支持分页和过滤\n - 输入参数:\n - `query` (string): 搜索关键词\n - `count` (number, 可选): 每页结果数量(最大20)\n - `offset` (number, 可选): 分页偏移量(最大9)\n\n- **location_search**\n \n - 搜索地理位置相关信息（商家、餐厅等）\n - 输入参数:\n - `query` (string): 位置搜索关键词\n - `count` (number, 可选): 结果数量(最大20)\n - 无相关结果时自动切换到网络搜索\n\n#### Metaso搜索引擎\n\n- **search**\n \n - 执行网络搜索，支持多种模式\n - 输入参数:\n - `query` (string): 搜索关键词\n - `mode` (string, 可选): 搜索模式\n - `concise`: 简洁模式，回答简短精炼\n - `detail`: 深入模式，回答详细全面（默认）\n - `research`: 研究模式，回答深度分析（目前暂不支持，逆向没有成功）\n\n- **scholar_search**\n \n - 执行学术搜索，专门用于查找学术资源\n - 输入参数:\n - `query` (string): 学术搜索关键词\n - `mode` (string, 可选): 搜索模式，同上\n\n#### 博查搜索引擎\n\n- **search**\n - 执行网络搜索，支持时间范围过滤和详细摘要\n - 输入参数:\n - `query` (string): 搜索关键词\n - `count` (number, 可选): 结果数量(1-10，默认10)\n - `page` (number, 可选): 页码，从1开始\n - `freshness` (string, 可选): 时间范围\n - `noLimit`: 不限时间（默认）\n - `oneDay`: 一天内\n - `oneWeek`: 一周内\n - `oneMonth`: 一月内\n - `oneYear`: 一年内\n - `summary` (boolean, 可选): 是否显示详细摘要，默认false\n - 返回内容:\n - 搜索统计信息（总结果数、当前页/总页数、本页结果数）\n - 网页搜索结果（标题、URL、来源、摘要、发布时间）\n - 相关图片信息（尺寸、来源、URL）\n\n## ","url","https://github.com/fengin/search-server","githubRepository",{"_76":77,"_78":79,"_80":81,"_82":83,"_84":85,"_86":87,"_91":92,"_93":94,"_96":97},"defaultBranch","main","fullName","fengin/search-server","githubTreeId",74624,"language","Python","name","search-server","owner",{"_88":89,"_17":90},"login","fengin","oilpxca8nw","recentStargazers",5,"spdxLicense",{"_84":95},"MIT License","stargazers",52,"glamaJson","id",945,"inspectable","integrations",[104],{"_105":106,"_39":112,"_17":113},"brand",{"_107":108,"_84":109,"_110":111},"logoSvgPath","M15.68 0l2.096 2.38s1.84-.512 2.709.358c.868.87 1.584 1.638 1.584 1.638l-.562 1.381.715 2.047s-2.104 7.98-2.35 8.955c-.486 1.919-.818 2.66-2.198 3.633-1.38.972-3.884 2.66-4.293 2.916-.409.256-.92.692-1.38.692-.46 0-.97-.436-1.38-.692a185.796 185.796 0 01-4.293-2.916c-1.38-.973-1.712-1.714-2.197-3.633-.247-.975-2.351-8.955-2.351-8.955l.715-2.047-.562-1.381s.716-.768 1.585-1.638c.868-.87 2.708-.358 2.708-.358L8.321 0h7.36zm-3.679 14.936c-.14 0-1.038.317-1.758.69-.72.373-1.242.637-1.409.742-.167.104-.065.301.087.409.152.107 2.194 1.69 2.393 1.866.198.175.489.464.687.464.198 0 .49-.29.688-.464.198-.175 2.24-1.759 2.392-1.866.152-.108.254-.305.087-.41-.167-.104-.689-.368-1.41-.741-.72-.373-1.617-.69-1.757-.69zm0-11.278s-.409.001-1.022.206-1.278.46-1.584.46c-.307 0-2.581-.434-2.581-.434S4.119 7.152 4.119 7.849c0 .697.339.881.68 1.243l2.02 2.149c.192.203.59.511.356 1.066-.235.555-.58 1.26-.196 1.977.384.716 1.042 1.194 1.464 1.115.421-.08 1.412-.598 1.776-.834.364-.237 1.518-1.19 1.518-1.554 0-.365-1.193-1.02-1.413-1.168-.22-.15-1.226-.725-1.247-.95-.02-.227-.012-.293.284-.851.297-.559.831-1.304.742-1.8-.089-.495-.95-.753-1.565-.986-.615-.232-1.799-.671-1.947-.74-.148-.068-.11-.133.339-.175.448-.043 1.719-.212 2.292-.052.573.16 1.552.403 1.632.532.079.13.149.134.067.579-.081.445-.5 2.581-.541 2.96-.04.38-.12.63.288.724.409.094 1.097.256 1.333.256s.924-.162 1.333-.256c.408-.093.329-.344.288-.723-.04-.38-.46-2.516-.541-2.961-.082-.445-.012-.45.067-.579.08-.129 1.059-.372 1.632-.532.573-.16 1.845.009 2.292.052.449.042.487.107.339.175-.148.069-1.332.508-1.947.74-.615.233-1.476.49-1.565.986-.09.496.445 1.241.742 1.8.297.558.304.624.284.85-.02.226-1.026.802-1.247.95-.22.15-1.413.804-1.413 1.169 0 .364 1.154 1.317 1.518 1.554.364.236 1.355.755 1.776.834.422.079 1.08-.4 1.464-1.115.384-.716.039-1.422-.195-1.977-.235-.555.163-.863.355-1.066l2.02-2.149c.341-.362.68-.546.68-1.243 0-.697-2.695-3.96-2.695-3.96s-2.274.436-2.58.436c-.307 0-.972-.256-1.585-.461-.613-.205-1.022-.206-1.022-.206z","Brave","slug","brave","Provides integration with Brave Search API, allowing users to perform web and location searches through the Brave Search engine.","6k2foanyoa","Search MCP Server","namespace","npmPackage","prompts","redditPostForeignId","1injbby","relatedMcpServerIds",[122,123,124,125,126,127],548,631,766,720,264,274,"resources","scores",{"_131":132,"_133":-5,"_134":-5},"license",100,"quality","security","simpleIcon","supportedPlatforms",[138,139],"LINUX","WINDOWS","tools",[142],{"_84":143,"_39":144,"_145":146},"search","执行网页搜索，从全网搜索任何网页信息和网页链接。结果准确、摘要完整，更适合AI使用。支持以下特性：\n- 时间范围过滤\n- 显示详细摘要\n- 分页获取\n每次请求最多返回10个结果。（当前使用博查搜索API实现）","inputSchema",{"_41":59,"_57":147,"_35":149},[148],"query",{"_150":151,"_155":156,"_148":159,"_161":162,"_165":166},"page",{"_41":152,"_54":153,"_39":154},"number",1,"页码，从1开始","count",{"_41":152,"_54":157,"_39":158},10,"结果数量(1-10，默认10)",{"_41":42,"_39":160},"搜索查询内容","summary",{"_41":163,"_54":28,"_39":164},"boolean","是否显示详细摘要","freshness",{"_167":168,"_41":42,"_54":169,"_39":174},"enum",[169,170,171,172,173],"noLimit","oneDay","oneWeek","oneMonth","oneYear","时间范围(noLimit:不限, oneDay:一天内, oneWeek:一周内, oneMonth:一月内, oneYear:一年内)","x9pd5cqiyx","updatedAt",["D",1742492101494],"meta",[180,183,185,192,195,197,200,203,206,209,212,214,217,220,222,224,226,228,230],{"_181":182},"title","Search MCP Server | Glama",{"_184":66,"_84":39},"content",{"_186":187,"_188":189,"_190":191},"href","https://glama.ai/mcp/servers/@fengin/search-server","rel","canonical","tagName","link",{"_184":114,"_193":194},"property","og:title",{"_184":66,"_193":196},"og:description",{"_184":198,"_193":199},"https://glama.ai/logo.png","og:logo",{"_184":201,"_193":202},"https://glama.ai/mcp/servers/@fengin/search-server/og-image","og:image",{"_184":204,"_193":205},"630","og:image:height",{"_184":207,"_193":208},"1200","og:image:width",{"_184":210,"_193":211},"website","og:type",{"_184":187,"_193":213},"og:url",{"_184":215,"_193":216},"ChatGPT alternative for power users","og:site_name",{"_184":218,"_84":219},"summary_large_image","twitter:card",{"_184":215,"_84":221},"twitter:title",{"_184":66,"_84":223},"twitter:description",{"_184":201,"_84":225},"twitter:image",{"_184":204,"_84":227},"twitter:image:height",{"_184":207,"_84":229},"twitter:image:width",{"_231":232},"script:ld+json",{"_233":234,"_235":236},"@context","http://schema.org","@graph",[237],{"_233":238,"_239":240,"_241":242},"https://schema.org","@type","BreadcrumbList","itemListElement",[243,249,253,257],{"_239":244,"_245":246,"_84":247,"_248":153},"ListItem","item","https://glama.ai/","Glama","position",{"_239":244,"_245":250,"_84":251,"_248":252},"https://glama.ai/mcp","MCP",2,{"_239":244,"_245":254,"_84":255,"_248":256},"https://glama.ai/mcp/servers","Servers",3,{"_239":244,"_245":187,"_84":114,"_248":258},4,"routes/_public/mcp/servers/~namespace/~slug/_pages/reviews/_route",{"_26":-5,"_29":261,"_178":294,"_327":-5},{"_31":262,"_33":263,"_13":271,"_65":66,"_67":66,"_68":272,"_74":273,"_98":-5,"_99":100,"_101":28,"_102":276,"_84":114,"_115":89,"_116":-5,"_117":-5,"_118":119,"_120":279,"_128":-5,"_129":280,"_135":-5,"_110":85,"_136":281,"_140":282,"_17":175,"_176":293},["D",1739336405764],{"_35":264,"_57":270,"_41":59},{"_37":265,"_43":266,"_46":267,"_49":268,"_52":269},{"_39":40,"_41":42},{"_39":45,"_41":42},{"_39":48,"_41":42},{"_39":51,"_41":42},{"_54":55,"_39":56,"_41":42},[],[61,62,63,64],{"_70":71,"_72":73},{"_76":77,"_78":79,"_80":81,"_82":83,"_84":85,"_86":274,"_91":92,"_93":275,"_96":97},{"_88":89,"_17":90},{"_84":95},[277],{"_105":278,"_39":112,"_17":113},{"_107":108,"_84":109,"_110":111},[122,123,124,125,126,127],{"_131":132,"_133":-5,"_134":-5},[138,139],[283],{"_84":143,"_39":144,"_145":284},{"_41":59,"_57":285,"_35":286},[148],{"_150":287,"_155":288,"_148":289,"_161":290,"_165":291},{"_41":152,"_54":153,"_39":154},{"_41":152,"_54":157,"_39":158},{"_41":42,"_39":160},{"_41":163,"_54":28,"_39":164},{"_167":292,"_41":42,"_54":169,"_39":174},[169,170,171,172,173],["D",1742492101494],[295,297,298,300,302,303,304,305,306,307,308,309,310,311,312,313,314,315,316],{"_181":296},"Reviews | Search MCP Server | Glama",{"_184":66,"_84":39},{"_186":299,"_188":189,"_190":191},"https://glama.ai/mcp/servers/@fengin/search-server/reviews",{"_184":301,"_193":194},"Reviews | Search MCP Server",{"_184":66,"_193":196},{"_184":198,"_193":199},{"_184":201,"_193":202},{"_184":204,"_193":205},{"_184":207,"_193":208},{"_184":210,"_193":211},{"_184":299,"_193":213},{"_184":215,"_193":216},{"_184":218,"_84":219},{"_184":215,"_84":221},{"_184":66,"_84":223},{"_184":201,"_84":225},{"_184":204,"_84":227},{"_184":207,"_84":229},{"_231":317},{"_233":234,"_235":318},[319],{"_233":238,"_239":240,"_241":320},[321,322,323,324,325],{"_239":244,"_245":246,"_84":247,"_248":153},{"_239":244,"_245":250,"_84":251,"_248":252},{"_239":244,"_245":254,"_84":255,"_248":256},{"_239":244,"_245":187,"_84":114,"_248":258},{"_239":244,"_245":299,"_84":326,"_248":92},"Reviews","review","actionData","errors"]

Rate & Review