SharePoint MCP Server

DISCLAIMER: This project is not affiliated with, endorsed by, or related to Microsoft Corporation. SharePoint and Microsoft Graph API are trademarks of Microsoft Corporation. This is an independent, community-driven project.

SharePoint MCP Server is a Model Context Protocol (MCP) server that connects LLM applications such as Claude to your SharePoint site via the Microsoft Graph API. Use natural language to query documents, manage lists, upload files, and more — directly from your AI assistant.

---

Features

Category	Capability
Site	Get site information
Libraries	Browse document libraries, list folder contents
Documents	Read DOCX, PDF, XLSX, CSV, TXT; browse by path; get item metadata; upload files
Search	Full-text search across all site content
Lists	Create lists with AI-optimized schemas; create, update list items
Pages	Create modern pages and news posts
Provisioning	Create new SharePoint sites and advanced document libraries
Transport	stdio (local), SSE, streamable-http (web / Docker)

---

Prerequisites

• Python 3.10 or higher
• A SharePoint site with Microsoft 365
• An Azure AD application registration with the required Graph API permissions (see docs/auth_guide.md)

---

Quickstart

1. Clone and install

git clone https://github.com/DEmodoriGatsuO/sharepoint-mcp.git
cd sharepoint-mcp

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

pip install -r requirements.txt

2. Configure

cp .env.example .env
# Edit .env with your Azure AD credentials and SharePoint site URL

Required variables in .env:

Variable	Description
TENANT_ID	Azure AD tenant ID
CLIENT_ID	Azure AD application (client) ID
CLIENT_SECRET	Azure AD client secret
SITE_URL	SharePoint site URL (https://{tenant}.sharepoint.com/sites/{name})

3. Verify your setup (optional)

python config_checker.py   # Validate configuration
python auth-diagnostic.py  # Test authentication

4. Start the server

# stdio — default, for Claude Desktop / MCP Inspector
python server.py

# HTTP streamable-http — for web services and Copilot agents
python server.py --transport streamable-http --port 8000

# Docker
docker-compose up

---

Usage

Claude Desktop

Install the server into Claude Desktop:

mcp install server.py --name "SharePoint Assistant"

Or add it manually to claude_desktop_config.json:

{
  "mcpServers": {
    "sharepoint": {
      "command": "python",
      "args": ["/absolute/path/to/sharepoint-mcp/server.py"],
      "env": {
        "TENANT_ID": "...",
        "CLIENT_ID": "...",
        "CLIENT_SECRET": "...",
        "SITE_URL": "..."
      }
    }
  }
}

MCP Inspector (development)

mcp dev server.py

HTTP Server

# streamable-http (recommended for Copilot agents and web clients)
python server.py --transport streamable-http --host 0.0.0.0 --port 8000

# SSE
python server.py --transport sse --host 0.0.0.0 --port 8000

# Via environment variables
MCP_TRANSPORT=streamable-http MCP_PORT=8000 python server.py

Docker

# Build and start (defaults to streamable-http on port 8000)
docker-compose up

# Or run manually
docker build -t sharepoint-mcp .
docker run --env-file .env -p 8000:8000 sharepoint-mcp

---

Available Tools

The following MCP tools are exposed to the LLM:

Tool	Description
get_site_info	Get name, description, URL, and metadata of the SharePoint site
list_document_libraries	List all document libraries (drives) in the site
list_folder_contents	Browse files and folders within a document library by path
get_document_content	Read and parse DOCX, PDF, XLSX, CSV, or TXT files
get_document_by_path	Retrieve document content by file path
get_item_metadata	Get metadata for a file or folder
search_sharepoint	Full-text search across all content in the site
upload_document	Upload a file to a document library
create_list_item	Create a new item in a SharePoint list
update_list_item	Update an existing item in a SharePoint list
create_intelligent_list	Provision a list with an AI-optimized schema
create_advanced_document_library	Create a document library with rich metadata
create_modern_page	Publish a modern SharePoint page
create_news_post	Publish a news article to the site
create_sharepoint_site	Provision a new SharePoint team site

For detailed usage examples and example prompts, see docs/usage.md.

---

Monitoring and Troubleshooting

Logs

The server writes logs to stdout. Set DEBUG=True in .env to enable verbose logging.

Common Issues

Symptom	Resolution
Authentication failure	Run python auth-diagnostic.py to diagnose
Permission errors	Verify your Azure AD app has the required Graph API permissions
Token issues	Run python token-decoder.py to inspect token claims

---

Contributing

Contributions are welcome. Please open an issue first to discuss significant changes. See CONTRIBUTING.md for guidelines.

All contributions must pass the quality checks before merge:

black .       # Formatting
ruff check .  # Linting
pytest        # Tests

---

License

Released under the MIT License. See LICENSE for details.