SharePoint MCP Server
A Model Context Protocol server for SharePoint file and folder management using Microsoft Graph API. Enables listing, uploading, downloading, and converting documents with markdown-to-DOCX support.
README
SharePoint MCP Server
A TypeScript MCP (Model Context Protocol) server for SharePoint file and folder management via Microsoft Graph API.
Features
- 11 MCP tools for complete SharePoint file/folder management
- Two auth flows: Client Credentials (app-level) and On-Behalf-Of (user-level)
- Two transports: stdio (for Claude Desktop/CLI) and Streamable HTTP
- Resumable uploads for large files (>4MB)
- Content extraction from PDF, Word (.docx), Excel (.xlsx), and text files
- Markdown-to-DOCX conversion via Pandoc with optional style templates from SharePoint
Quick Start
1. Install dependencies
npm install
npm run build
2. Configure environment
cp .env.example .env
# Edit .env with your Azure AD and SharePoint details
Required environment variables:
AZURE_TENANT_ID- Azure AD tenant IDAZURE_CLIENT_ID- App registration client IDAZURE_CLIENT_SECRET- App registration client secretSHAREPOINT_HOSTNAME- e.g.yourcompany.sharepoint.comSHAREPOINT_SITE_NAME- e.g.YourSiteNameAUTH_FLOW-client_credentials(default) oron_behalf_of
3. Run
stdio transport (for Claude Desktop):
npm start
HTTP transport:
npm run start:http
# Server starts on http://localhost:3000/mcp
Claude Desktop Configuration
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"sharepoint": {
"command": "node",
"args": ["/path/to/sharepoint-mcp-nodejs/dist/index.js"],
"env": {
"AZURE_TENANT_ID": "your-tenant-id",
"AZURE_CLIENT_ID": "your-client-id",
"AZURE_CLIENT_SECRET": "your-client-secret",
"SHAREPOINT_HOSTNAME": "yourcompany.sharepoint.com",
"SHAREPOINT_SITE_NAME": "YourSiteName"
}
}
}
}
Available Tools
| Tool | Description |
|---|---|
List_SharePoint_Folders |
List folders in a directory |
Create_Folder |
Create a new folder |
Delete_Folder |
Delete a folder |
Get_SharePoint_Tree |
Recursive tree view (configurable depth) |
List_SharePoint_Documents |
List documents with metadata |
Get_Document_Content |
Download + extract text (PDF/Word/Excel/text) |
Create_Document_From_Markdown |
Convert markdown/text to .docx with optional style template |
Upload_Document |
Upload from base64 or text content |
Upload_Document_From_Path |
Upload a local file (resumable for >4MB) |
Update_Document |
Update existing document content |
Delete_Document |
Delete a document |
Markdown to DOCX Conversion
The Create_Document_From_Markdown tool converts markdown or plain text to a .docx file using Pandoc and uploads it to SharePoint.
Parameters:
content(required) — Markdown or plain text to convertfolder_path(required) — Destination folder in SharePointfile_name(required) — Output filename (e.g.report.docx)template_path(optional) — Path to a .docx template on SharePoint for styling
Style templates: Create a .docx file in Word with your desired styles (fonts, headings, margins, colors), upload it to SharePoint, then reference it as template_path. Pandoc applies styles from the template to the generated document via --reference-doc.
Testing the Markdown-to-DOCX Tool
Prerequisites: Pandoc must be installed. In Docker this is handled automatically. For local development:
# macOS
brew install pandoc
# Ubuntu/Debian
sudo apt-get install pandoc
# Verify
pandoc --version
Test 1 — Basic conversion (no template):
Using MCP Inspector:
npx @modelcontextprotocol/inspector node dist/index.js
Call Create_Document_From_Markdown with:
{
"content": "# My Report\n\n## Introduction\n\nThis is a **test** document with:\n\n- Bullet point 1\n- Bullet point 2\n\n## Data\n\n| Name | Value |\n|------|-------|\n| A | 100 |\n| B | 200 |",
"folder_path": "Shared Documents",
"file_name": "test-report.docx"
}
Then verify with Get_Document_Content:
{
"file_path": "Shared Documents/test-report.docx"
}
Test 2 — With style template:
- Create a styled .docx template in Word (set fonts, heading styles, margins, colors)
- Upload it to SharePoint (e.g. via
Upload_Document_From_Pathor manually) - Call
Create_Document_From_Markdownwith the template:
{
"content": "# Styled Report\n\nThis document uses corporate styling.",
"folder_path": "Shared Documents",
"file_name": "styled-report.docx",
"template_path": "Templates/corporate-style.docx"
}
Test 3 — Via curl (HTTP transport):
# Initialize session
curl -s -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"test","version":"0.1.0"}}}'
# Call the tool (use the mcp-session-id from the response headers above)
curl -s -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-H "mcp-session-id: <SESSION_ID>" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"Create_Document_From_Markdown","arguments":{"content":"# Hello World\n\nTest document.","folder_path":"Shared Documents","file_name":"curl-test.docx"}}}'
Upload Strategy
- Files <= 4MB: Simple PUT to
/contentendpoint - Files > 4MB: Resumable upload session with 3.2MB chunks
Auth Flows
Client Credentials (default)
App-level access using client ID + secret. Best for server-to-server scenarios.
On-Behalf-Of (OBO)
User-level access by exchanging a user token. Use with HTTP transport where the client sends a Authorization: Bearer <token> header.
See docs/azure-setup.md for detailed Azure Portal setup instructions.
Docker
After editing .env with your credentials, build and run with:
docker compose up --build -d
The server will be available at http://localhost:3000/mcp.
After making changes to the source, rebuild and restart (the Docker build compiles TypeScript internally):
docker compose up --build -d
To view logs:
docker compose logs -f
Development
npm run dev # Watch mode for TypeScript compilation
Testing with MCP Inspector
npx @modelcontextprotocol/inspector node dist/index.js
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。