SharePoint MCP Server and AI Chatbot

This project provides a Model Context Protocol (MCP) server for SharePoint Online and an AI-powered chatbot that can interact with SharePoint content.

Features

• List available SharePoint documents and pages
• Fetch and read document content from SharePoint
• Query SharePoint pages (ASPX)
• Search SharePoint content
• Summarize document paragraphs
• Convert Office documents to text
• Interactive AI chatbot interface

Prerequisites

• Node.js 16+ or Bun runtime
• SharePoint Online site
• Microsoft Graph API credentials
• OpenAI API key (for AI chatbot)

Setup

1. Install dependencies:

   bun install

2. Configure environment variables:

   # SharePoint/Microsoft Graph credentials
   export TENANT_ID="your-tenant-id"
   export CLIENT_ID="your-client-id"
   export CLIENT_SECRET="your-client-secret"
   export SITE_ID="your-sharepoint-site-id"
      
   # For AI Chatbot
   export OPENAI_API_KEY="your-openai-api-key"

Running the Application

IMPORTANT: Two-Step Process

The AI chatbot connects to the MCP server over HTTP, so you need to run them as separate processes:

Step 1: Start the HTTP server first:

bun run http

Step 2: In a separate terminal, start the AI chatbot:

bun run ai

Other Run Options

• Command-line chatbot (no AI):

  bun run cli

• Run MCP server with stdio transport (for direct integration):

  bun run stdio

Available Commands in AI Chatbot

• list documents - Show all available SharePoint documents
• list pages - Show all SharePoint pages
• search [query] - Search for content in SharePoint
• document [id] - Get content of a specific document
• page [id] - Get content of a specific SharePoint page
• summarize [id] - Get a summary of document content
• exit or quit - Exit the chatbot

Troubleshooting

• "No valid session ID provided" error: Make sure you're running the HTTP server (Step 1) before starting the AI chatbot (Step 2).
• OpenAI API errors: Check that your OPENAI_API_KEY is set correctly.
• SharePoint connection issues: Verify your Microsoft Graph credentials and site ID.

License

MIT

Available Resources

| Resource URI | Description | |--------------|-------------| | sharepoint://documents | Lists all documents in the root folder | | sharepoint://document/{documentId} | Retrieves a specific document's content | | sharepoint://pages | Lists all SharePoint pages | | sharepoint://page/{pageId} | Retrieves a specific page's content |

Available Tools

| Tool Name | Parameters | Description | |-----------|------------|-------------| | search-sharepoint | { query: string } | Searches SharePoint content | | summarize-content | { documentId: string, paragraphIndex?: number } | Summarizes document content or specific paragraphs | | document-to-text | { documentId: string } | Converts Office documents to text |

Azure App Registration

1. Register a new application in the Azure Portal
2. Add the following API permissions:
   • Microsoft Graph > Application permissions:
     • Files.Read.All
     • Sites.Read.All
   • SharePoint > Application permissions:
     • Sites.Read.All
3. Generate a client secret
4. Grant admin consent for the permissions

Finding Your SharePoint Site ID

You can find your SharePoint site ID using the Microsoft Graph Explorer:

1. Go to https://developer.microsoft.com/graph/graph-explorer
2. Sign in with your Microsoft 365 account
3. Make a GET request to: /sites/{domain}:/sites/{site-name} Example: /sites/contoso.sharepoint.com:/sites/marketing
4. The response will include your site ID

Usage with AI Models

Example: Using with a model like Claude or GPT

1. Connect your AI client to this MCP server
2. List available documents:
   • Request the sharepoint://documents resource
3. Read a specific document:
   • Request the sharepoint://document/{documentId} resource
4. Search for content:
   • Call the search-sharepoint tool with a query parameter
5. Summarize a document:
   • Call the summarize-content tool with a documentId parameter

Example: Flow for Document Summarization

1. AI: "I need to find documents about marketing strategy"
2. User: "Let me help you search SharePoint" 
3. [AI calls search-sharepoint tool with query="marketing strategy"]
4. AI: "I found a marketing plan document. Let me get its content"
5. [AI requests the sharepoint://document/{documentId} resource]
6. AI: "Here's a summary of the marketing plan..."
7. User: "Can you focus on the first quarter strategy?"
8. [AI calls summarize-content tool with documentId and paragraphIndex parameters]

Architecture

This MCP server uses:

• Microsoft Graph API: To interact with SharePoint Online
• Azure Identity: For authentication with Microsoft services
• MCP SDK: To implement the Model Context Protocol
• Two transport options:
  • StdioServerTransport: For direct integration with AI clients
  • StreamableHTTPServerTransport: For remote access over HTTP

Extending the Server

You can extend this server by:

1. Adding more resources in index.ts or http-server.ts
2. Implementing additional tools for specific SharePoint operations
3. Supporting more document types and extraction methods