📧 Gmail MCP Project

MCP Server and MCP Client created with FastMCP and Python for Gmail and OpenAI integration.

A complete Model Context Protocol (MCP) implementation that enables AI assistants to interact with Gmail through a FastMCP server, with a Streamlit web interface for easy interaction.

🌟 Features

• MCP Server - FastMCP-based server exposing Gmail operations
• MCP Client - Python client integrating OpenAI GPT-4o-mini with Gmail toolsç
• MCP Client - also with Ollama Now
• Streamlit Frontend - Web interface for the MCP client
• Gmail Integration - List, send, and manage emails
• Resource Templates - Dynamic PDF manual access
• Prompt Templates - Pre-built workflows for email automation

🏗️ Architecture

1. MCP Server (gmail_mcp_server.py)

The server exposes Gmail functionality through the MCP protocol:

• Tools: list_emails, send_email
• Resources: Gmail profile information
• Resource Templates: PDF manual access with versioning
• Prompts: Email summarization, professional email composition, automation workflows

2. MCP Client (client.py) or (client_ollama.py)

The client bridges OpenAI and the MCP server:

• Connects to the MCP server using FastMCP Client
• Converts MCP tools to OpenAI function calling format or Ollama format
• Handles chat completions with tool execution
• Manages resources and prompts

3. Streamlit App (app.py)

Frontend interface providing:

• Interactive chat with the Gmail assistant
• Quick access to prompt templates
• System information display
• Real-time email management

📋 Prerequisites

• Python 3.10+
• OpenAI API Key - Get from OpenAI Platform
• Google Cloud Credentials - Create OAuth 2.0 credentials in Google Cloud Console
• Ollama - get From (https://ollama.com/)

🚀 Installation

1. Create Virtual Environment

Using venv:

python -m venv .venv
.venv\Scripts\activate  # Windows
source .venv/bin/activate  # Linux/Mac

2. Install Dependencies

pip install openai fastmcp python-dotenv
pip install streamlit
pip install google-auth-oauthlib google-api-python-client PyPDF2
pip install ollama

⚙️ Configuration

1. OpenAI API Key

Create a .env file in the project root:

OPENAI_API_KEY=sk-your-openai-api-key-here

Important: Add .env to .gitignore to keep your API key secure.

2. Google Gmail Credentials

1. Go to Google Cloud Console
2. Create a new project or select an existing one
3. Enable the Gmail API
4. Create OAuth 2.0 Client ID credentials (Desktop application)
5. Download the credentials and save as credentials.json in the project root

Important: Add credentials.json and token.pickle to .gitignore.

3. File Structure

.
├── app.py                    # Streamlit frontend
├── client.py                 # MCP Client
├── gmail_mcp_server.py       # MCP Server
├── credentials.json          # Google OAuth credentials (not in git)
├── .env                      # OpenAI API key (not in git)
├── token.pickle             # Gmail auth token (auto-generated, not in git)
├── manuals/                  # PDF manuals directory
└── README.md

🎯 Usage

Run the Streamlit App

streamlit run app.py

Or with the full virtual environment path:

".venv/Scripts/streamlit.exe" run app.py  # Windows
.venv/bin/streamlit run app.py            # Linux/Mac

Run the MCP Server Standalone

python gmail_mcp_server.py

First Run - Gmail Authentication

On first execution, the app will:

1. Open your browser for Google authentication
2. Request permissions to read and send emails
3. Save authentication token to token.pickle

💡 How It Works

Client Workflow

1. OpenAI Integration

   • Client connects to OpenAI using the API key from .env
   • Uses GPT-4o-mini model for intelligent responses

2. Gmail Connection

   • Client loads credentials.json for Google authentication
   • Manages OAuth flow and token refresh automatically

3. FastMCP Client Methods

   • list_tools() - Retrieves available MCP tools
   • call_tool() - Executes tools on the MCP server
   • get_prompt() - Loads prompt templates
   • read_resource() - Accesses resource data

4. OpenAI Tool Format Conversion

   • Converts MCP tools to OpenAI's function calling JSON structure
   • Each LLM has specific JSON requirements
   • Client handles translation between MCP and OpenAI formats

📚 Available Commands

List Recent Emails

"Show me my last 10 emails"

Send an Email

"Send an email to john@example.com with subject 'Meeting' and body 'Let's meet tomorrow'"

Use Prompt Templates

• Daily Email Summary - Get organized overview of today's emails
• Compose Professional Email - AI-assisted email writing

🔧 Troubleshooting

Import Errors

Make sure you're using the virtual environment Python:

which python  # Should show .venv path

Google Authentication Issues

Delete token.pickle and authenticate again:

del token.pickle  # Windows
rm token.pickle   # Linux/Mac

OpenAI API Errors

Verify your .env file contains a valid API key:

cat .env  # Should show OPENAI_API_KEY=sk-...

📝 License

This project is for educational purposes.

👨‍💻 Author JCDiaz

Created with FastMCP and OpenAI integration.