Outlook MCP Server

A comprehensive Model Context Protocol (MCP) server that provides secure, efficient access to Microsoft Outlook functionality through Python COM automation. This server enables AI assistants and applications to interact with Outlook emails, folders, and contacts programmatically while maintaining enterprise-grade security and performance.

🚀 Overview

The Outlook MCP Server bridges the gap between AI systems and Microsoft Outlook, providing a standardized interface for email operations. Built on the Model Context Protocol, it offers both programmatic API access and interactive CLI usage patterns.

Key Capabilities

• Email Operations: Search, retrieve, compose, and reply to emails
• Folder Management: Browse and access all Outlook folders
• Batch Processing: Send bulk emails with CSV-based recipient lists (CLI only)
• Caching System: Intelligent email caching for performance optimization
• Security: Built-in user confirmation for email sending operations
• Error Handling: Comprehensive error handling with retry mechanisms

📋 Requirements

System Requirements

• Operating System: Windows 10/11 (required for Outlook COM automation)
• Python: 3.8 or higher
• Microsoft Outlook: 2016 or later, properly configured and running
• COM Access: Outlook must be accessible via COM (default for most installations)

Dependencies

• fastmcp: MCP server framework
• pywin32: Windows COM automation
• Standard library: argparse, csv, datetime, logging, typing

🛠️ Installation

Quick Start

# Clone the repository
git clone https://github.com/marlonluo2018/outlook-mcp-server.git
cd outlook-mcp-server

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install requirements
pip install -r requirements.txt

MCP Configuration

Add to your MCP client configuration (e.g., Claude Desktop settings.json):

{
  "mcpServers": {
    "outlook": {
      "type": "stdio",
      "command": "python",
      "args": ["C:\\Project\\outlook-mcp-server\\outlook_mcp_server.py"]
    }
  }
}

Human Interface: CLI

The CLI interface is designed exclusively for human users, not LLMs:

• Purpose: Provides a human-friendly interactive interface
• Audience: Only for direct human operation
• Security: All operations require interactive confirmation
• Workflow:
  1. Load emails into cache (List/Search)
  2. Operate on cached emails (View/Reply/Compose)
  3. Cache auto-refreshes on new operations

# Start interactive session (human only)
python cli_interface.py

Note: For LLM integration, use the MCP server interface instead.

Development Installation

# Install development dependencies
pip install -e ".[dev]"

Environment Variables

# Optional: Set log level
export OUTLOOK_MCP_LOG_LEVEL=INFO  # DEBUG, INFO, WARNING, ERROR

# Optional: Set cache timeout
export OUTLOOK_MCP_CACHE_TIMEOUT=300  # seconds

Configuration Constants

Located in backend/shared.py:

Constant	Default	Description
MAX_DAYS	30	Maximum days to look back for emails
MAX_EMAILS	1000	Maximum emails to process in one operation
MAX_LOAD_TIME	58	Maximum processing time per operation (seconds)
CONNECT_TIMEOUT	30	Connection timeout for Outlook
MAX_RETRIES	3	Maximum retry attempts for failed operations

🎯 Usage

MCP Server Mode (For LLM Tool Calls)

Tool Call Sequence:

1. First call either:
   • list_recent_emails_tool to load recent emails into cache
   • search_emails_tool to load matching emails into cache
2. Then call view_email_cache_tool to identify specific email to work with
3. LLM drafts reply content based on cached email
4. Finally call reply_to_email_by_number_tool (requires user confirmation)

Key Points:

• All operations work with the email cache
• Write operations require explicit user confirmation
• Cache is automatically refreshed on new list/search

CLI Interface Workflow

The CLI interface follows a strict email cache workflow:

1. Cache Population:

   • List or search operations load emails into memory cache
   • Cache contains email metadata and partial content

2. Cache Operations:

   • View full email details from cache
   • Reply to cached emails
   • Use cached emails as templates

3. Cache Management:

   • Cache automatically refreshes on new list/search
   • In-memory only - no persistent storage
   • Limited to most recent 1000 emails

# Start interactive session
python cli_interface.py

Available Tools

1. List Folders

{
  "tool": "get_folder_list_tool",
  "parameters": {}
}
// Returns: {"content": [{"type": "text", "text": "['Inbox', 'Sent Items', 'Drafts', ...]"}]}

2. List Recent Emails

{
  "tool": "list_recent_emails_tool",
  "parameters": {
    "days": 7,
    "folder_name": "Inbox"
  }
}
// Returns count and first page preview

3. Search Emails

{
  "tool": "search_emails_tool",
  "parameters": {
    "search_term": "meeting notes",
    "days": 14,
    "folder_name": "Inbox",
    "match_all": true
  }
}

4. View Email Cache

{
  "tool": "view_email_cache_tool",
  "parameters": {
    "page": 1
  }
}

5. Get Email Details

{
  "tool": "get_email_by_number_tool",
  "parameters": {
    "email_number": 3
  }
}
// Returns full email with body and attachments

6. Reply to Email

{
  "tool": "reply_to_email_by_number_tool",
  "parameters": {
    "email_number": 5,
    "reply_text": "Thank you for your message...",
    "to_recipients": ["custom@example.com"],
    "cc_recipients": ["boss@example.com"]
  }
}
// ⚠️ Requires explicit user confirmation

7. Compose New Email

{
  "tool": "compose_email_tool",
  "parameters": {
    "recipient_email": "client@example.com",
    "subject": "Project Update",
    "body": "Dear team,\n\nHere's the latest update...",
    "cc_email": "manager@example.com"
  }
}
// ⚠️ Requires explicit user confirmation

8. Batch Email Operations (Interactive Only)

Workflow:

1. First load template emails into cache via List/Search
2. Select cached email as template in interactive mode
3. Provide CSV of recipients and optional custom text
4. Confirm before sending

Note: Batch operations require working with the email cache and are only available through interactive CLI.

📊 Data Flow Architecture

Email Processing Pipeline

Outlook COM → Session Manager → Email Parser → Cache → API Response

Cache System

• Key: Email EntryID (unique Outlook identifier)
• Value: Structured email data (subject, sender, body, attachments)
• Lifetime: Cache cleared on new email listing/search
• Format: JSON-serializable Python dictionaries

Error Handling Flow

Operation → Try/Catch → Retry Logic → User-friendly Error → Logging

🔒 Security Considerations

Email Sending Protection

• Explicit Confirmation: All email sending operations require user approval
• Rate Limiting: Batch operations limited to 500 recipients per batch
• Input Validation: All inputs sanitized and validated before processing

Data Privacy

• Local Processing: All operations performed locally on user's machine
• No External Calls: No data transmitted outside Outlook COM interface
• Cache Isolation: Email cache stored in memory only, no persistent storage

📈 Performance Optimization

Caching Strategy

• Smart Loading: Only load requested email ranges
• Batch Processing: Process emails in configurable batches
• Timeout Protection: Automatic termination of long-running operations

Memory Management

• Streaming: Large email bodies truncated in cache
• Garbage Collection: Automatic cleanup of COM objects
• Resource Limits: Configurable maximum email processing limits

🤝 Contributing

Development Setup

# Fork and clone
git clone https://github.com/marlonluo2018/outlook-mcp-server.git
cd outlook-mcp-server

# Install development dependencies
pip install -e ".[dev]"



### Pull Request Guidelines
1. **Tests**: Add tests for new functionality
2. **Documentation**: Update README for API changes
3. **Error Handling**: Include proper error handling
4. **Performance**: Consider impact on large email volumes
5. **Security**: Validate all user inputs

### Code Style
- **PEP 8**: Follow Python style guidelines
- **Type Hints**: Use typing module for all functions
- **Docstrings**: Google-style docstrings for all public APIs
- **Error Messages**: User-friendly error messages

## 📋 Changelog

### v1.0.0 (Current)
- **Initial Release**: Complete MCP server implementation
- **Core Features**: Email retrieval, search, composition, and batch operations
- **Security**: User confirmation for email sending
- **Performance**: Email caching and timeout handling
- **CLI**: Interactive and command-line interfaces

### Planned Features
- [ ] Attachment upload/download support
- [ ] Calendar integration
- [ ] Contact management
- [ ] Email rules/filters
- [ ] Multi-account support
- [ ] Webhook notifications

## 📄 License

MIT License - see [LICENSE](LICENSE) file for details.

## 🆘 Support

- **Issues**: [GitHub Issues](https://github.com/marlonluo2018/outlook-mcp-server/issues)
- **Discussions**: [GitHub Discussions](https://github.com/marlonluo2018/outlook-mcp-server/discussions)
- **Documentation**: This README and inline code documentation