概要
What is Gmail MCP Server?
Gmail MCP Server is a FastMCP-based integration that exposes Gmail operations as tools for AI assistants. It uses Google OAuth 2.0 and the Gmail API to send, read, manage, and trash emails. The project also includes a smart client that leverages Google’s Generative AI (Gemini) to interpret natural language requests.
How to use Gmail MCP Server?
Install dependencies from requirements.txt, place Gmail API OAuth credentials as credentials.json, and set the Gemini API key in a .env file. Start the server with python gmail_server.py --creds credentials.json --token token.json. Then run the smart client with python test_gmail_server.py to interact using natural language commands.
Key features of Gmail MCP Server
- Exposes tools:
send_email,get_unread_emails,read_email,trash_email,open_email,mark_email_as_read - Includes prompts for email drafting and management
- LLM‑powered client understands plain‑English requests
- Handles OAuth authentication and automatic token refresh
- Parses MIME messages and decodes email headers
- Comprehensive logging and error handling
Use cases of Gmail MCP Server
- Send an email to a recipient about a specific topic
- Check and read unread messages
- Open the latest email in the browser
- Move an email from a sender to trash
- Mark all unread emails as read
FAQ from Gmail MCP Server
What are the prerequisites for using Gmail MCP Server?
Python 3.6+, Gmail API credentials from Google Cloud Console, a Gemini API key from Google AI Studio, and the packages listed in requirements.txt.
How does authentication work?
The server uses OAuth 2.0. On first run, it opens a browser for you to log in to Google and grant permissions, then saves a token.json file for subsequent requests.
Where are authentication and configuration files stored?
credentials.json (OAuth client ID), token.json (generated token), and .env (Gemini API key) reside in the project root. The .gitignore prevents accidental commits of these sensitive files.
How can I add new Gmail operations?
Add the tool implementation in gmail_server.py. The smart client automatically discovers and exposes new tools, and the LLM learns to handle requests from the tool descriptions.
Does Gmail MCP Server include a natural language interface?
Yes. The test_gmail_server.py client uses Gemini to interpret plain‑English requests (e.g., “Send an email to Alice”) and converts them into the appropriate tool calls, asking for missing information when needed.
