饺子AI MCP服务器
一种模型上下文协议（MCP）服务器实现，与Dumpling AI集成，用于数据抓取、内容处理、知识管理、AI代理和代码执行功能。
特性
- 与Dumpling AI API所有端点完全集成
- 用于YouTube转录、搜索、自动补全、地图、地点、新闻和评论的数据API
- 网络抓取，支持抓取、爬行、截图和结构化数据提取
- 用于文本提取、PDF操作、视频处理的文档转换工具
- 从文档、图像、音频和视频中提取数据
- 人工智能功能，包括代理完成、知识库管理和图像生成
- 用于在安全环境中运行JavaScript和Python代码的开发工具
- 自动错误处理和详细的响应格式
安装
通过Smithery安装
通过以下方式自动安装Claude Desktop的mcp服务器dumplingai 史密瑟里:
npx -y @smithery/cli install @Dumpling-AI/mcp-server-dumplingai --client claude使用npx运行
env DUMPLING_API_KEY=your_api_key npx -y mcp-server-dumplingai手动安装
npm install -g mcp-server-dumplingai在游标上运行
配置光标🖥️ 注意：需要游标版本0.45.6+
要在游标中配置饺子AI MCP：
- 打开光标设置
- 转到功能>MCP服务器
- 点击“+添加新MCP服务器”
- 输入以下内容：
{
"mcpServers": {
"dumplingai": {
"command": "npx",
"args": ["-y", "mcp-server-dumplingai"],
"env": {
"DUMPLING_API_KEY": ""
}
}
}
}如果您正在使用Windows并遇到问题，请尝试 cmd /c "set DUMPLING_API_KEY=your-api-key && npx -y mcp-server-dumplingai"替换 your-api-key 使用Dumpling AI API密钥。
配置
环境变量
DUMPLING_API_KEY：您的Dumpling AI API密钥（必需）
可用工具
数据API
1.获取YouTube成绩单(get-youtube-transcript)
从带有可选时间戳的YouTube视频中提取转录本。
{
"name": "get-youtube-transcript",
"arguments": {
"videoUrl": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
"includeTimestamps": true,
"timestampsToCombine": 3,
"preferredLanguage": "en"
}
}2.搜索(search)
执行谷歌网络搜索，并可选择从结果中抓取内容。
{
"name": "search",
"arguments": {
"query": "machine learning basics",
"country": "us",
"language": "en",
"dateRange": "pastMonth",
"scrapeResults": true,
"numResultsToScrape": 3,
"scrapeOptions": {
"format": "markdown",
"cleaned": true
}
}
}3.获取自动补全功能(get-autocomplete)
获取查询的谷歌搜索自动完成建议。
{
"name": "get-autocomplete",
"arguments": {
"query": "how to learn",
"country": "us",
"language": "en",
"location": "New York"
}
}4.搜索地图(search-maps)
在谷歌地图上搜索地点和企业。
{
"name": "search-maps",
"arguments": {
"query": "coffee shops",
"gpsPositionZoom": "37.7749,-122.4194,14z",
"language": "en",
"page": 1
}
}5.搜索地点(search-places)
搜索有更详细信息的地方。
{
"name": "search-places",
"arguments": {
"query": "hotels in paris",
"country": "fr",
"language": "en",
"page": 1
}
}6.搜索新闻(search-news)
搜索具有可自定义参数的新闻文章。
{
"name": "search-news",
"arguments": {
"query": "climate change",
"country": "us",
"language": "en",
"dateRange": "pastWeek"
}
}7.获取谷歌评论(get-google-reviews)
检索企业或地点的谷歌评论。
{
"name": "get-google-reviews",
"arguments": {
"businessName": "Eiffel Tower",
"location": "Paris, France",
"limit": 10,
"sortBy": "relevance"
}
}Web剪贴
8.废料(scrape)
使用格式选项从网页中提取内容。
{
"name": "scrape",
"arguments": {
"url": "https://example.com",
"format": "markdown",
"cleaned": true,
"renderJs": true
}
}9.爬行(crawl)
递归抓取网站并使用可定制的参数提取内容。
{
"name": "crawl",
"arguments": {
"baseUrl": "https://example.com",
"maxPages": 10,
"crawlBeyondBaseUrl": false,
"depth": 2,
"scrapeOptions": {
"format": "markdown",
"cleaned": true,
"renderJs": true
}
}
}10.屏幕截图(screenshot)
使用可自定义的视口和格式选项捕获网页的屏幕截图。
{
"name": "screenshot",
"arguments": {
"url": "https://example.com",
"width": 1280,
"height": 800,
"fullPage": true,
"format": "png",
"waitFor": 1000
}
}11.提取(extract)
使用人工智能指令从网页中提取结构化数据。
{
"name": "extract",
"arguments": {
"url": "https://example.com/products",
"instructions": "Extract all product names, prices, and descriptions from this page",
"schema": {
"products": [
{
"name": "string",
"price": "number",
"description": "string"
}
]
},
"renderJs": true
}
}文档转换
12.文件转文本(doc-to-text)
使用可选的OCR将文档转换为明文。
{
"name": "doc-to-text",
"arguments": {
"url": "https://example.com/document.pdf",
"options": {
"ocr": true,
"language": "en"
}
}
}13.转换为PDF(convert-to-pdf)
将各种文件格式转换为PDF。
{
"name": "convert-to-pdf",
"arguments": {
"url": "https://example.com/document.docx",
"format": "docx",
"options": {
"quality": 90,
"pageSize": "A4",
"margin": 10
}
}
}14.合并PDF(merge-pdfs)
将多个PDF合并到一个文档中。
{
"name": "merge-pdfs",
"arguments": {
"urls": ["https://example.com/doc1.pdf", "https://example.com/doc2.pdf"],
"options": {
"addPageNumbers": true,
"addTableOfContents": true
}
}
}15.修剪视频(trim-video)
从视频中提取特定片段。
{
"name": "trim-video",
"arguments": {
"url": "https://example.com/video.mp4",
"startTime": 30,
"endTime": 60,
"output": "mp4",
"options": {
"quality": 720,
"fps": 30
}
}
}16.提取文件(extract-document)
从各种格式的文档中提取特定内容。
{
"name": "extract-document",
"arguments": {
"url": "https://example.com/document.pdf",
"format": "structured",
"options": {
"ocr": true,
"language": "en",
"includeMetadata": true
}
}
}17.提取图像(extract-image)
从图像中提取文本和信息。
{
"name": "extract-image",
"arguments": {
"url": "",
"extractionType": "text",
"options": {
"language": "en",
"detectOrientation": true
}
}
}18.提取音频(extract-audio)
转录和提取音频文件中的信息。
{
"name": "extract-audio",
"arguments": {
"url": "https://example.com/audio.mp3",
"language": "en",
"options": {
"model": "enhanced",
"speakerDiarization": true,
"wordTimestamps": true
}
}
}19.提取视频(extract-video)
从视频中提取内容，包括文字记录、场景和对象。
{
"name": "extract-video",
"arguments": {
"url": "https://example.com/video.mp4",
"extractionType": "transcript",
"options": {
"language": "en",
"speakerDiarization": true
}
}
}20.阅读PDF元数据(read-pdf-metadata)
从PDF文件中提取元数据。
{
"name": "read-pdf-metadata",
"arguments": {
"url": "https://example.com/document.pdf",
"includeExtended": true
}
}21.编写PDF元数据(write-pdf-metadata)
更新PDF文件中的元数据。
{
"name": "write-pdf-metadata",
"arguments": {
"url": "https://example.com/document.pdf",
"metadata": {
"title": "New Title",
"author": "John Doe",
"keywords": ["keyword1", "keyword2"]
}
}
}人工智能
22.生成代理完成(generate-agent-completion)
使用可选工具定义获取AI代理完成。
{
"name": "generate-agent-completion",
"arguments": {
"prompt": "How can I improve my website's SEO?",
"model": "gpt-4",
"temperature": 0.7,
"maxTokens": 500,
"context": ["The website is an e-commerce store selling handmade crafts."]
}
}23.搜索知识库(search-knowledge-base)
在知识库中搜索相关信息。
{
"name": "search-knowledge-base",
"arguments": {
"kbId": "kb_12345",
"query": "How to optimize database performance",
"limit": 5,
"similarityThreshold": 0.7
}
}24.添加到知识库(add-to-knowledge-base)
向知识库添加条目。
{
"name": "add-to-knowledge-base",
"arguments": {
"kbId": "kb_12345",
"entries": [
{
"text": "MongoDB is a document-based NoSQL database.",
"metadata": {
"source": "MongoDB documentation",
"category": "databases"
}
}
],
"upsert": true
}
}25.生成AI图像(generate-ai-image)
使用AI模型生成图像。
{
"name": "generate-ai-image",
"arguments": {
"prompt": "A futuristic city with flying cars and neon lights",
"width": 1024,
"height": 1024,
"numImages": 1,
"quality": "hd",
"style": "photorealistic"
}
}26.生成图像(generate-image)
使用各种AI提供商生成图像。
{
"name": "generate-image",
"arguments": {
"prompt": "A golden retriever in a meadow of wildflowers",
"provider": "dalle",
"width": 1024,
"height": 1024,
"numImages": 1
}
}开发者工具
27.运行JavaScript代码(run-js-code)
执行具有可选依赖性的JavaScript代码。
{
"name": "run-js-code",
"arguments": {
"code": "const result = [1, 2, 3, 4].reduce((sum, num) => sum + num, 0); console.log(`Sum: ${result}`); return result;",
"dependencies": {
"lodash": "^4.17.21"
},
"timeout": 5000
}
}28.运行Python代码(run-python-code)
使用可选依赖性执行 Python 代码。
{
"name": "run-python-code",
"arguments": {
"code": "import numpy as np\narr = np.array([1, 2, 3, 4, 5])\nmean = np.mean(arr)\nprint(f'Mean: {mean}')\nreturn mean",
"dependencies": ["numpy", "pandas"],
"timeout": 10000,
"saveOutputFiles": true
}
}错误处理
服务器提供强大的错误处理：
- 带有HTTP状态代码的详细错误消息
- API密钥验证
- 使用Zod模式进行输入验证
- 使用描述性消息处理网络错误
错误响应示例：
{
"content": [
{
"type": "text",
"text": "Error: Failed to fetch YouTube transcript: 404 Not Found"
}
],
"isError": true
}发展
# Install dependencies
npm install
# Build
npm run build
许可证
MIT许可证-有关详细信息，请参阅许可证文件