VulcanRaven737/IEEE-CS-GirlGeeks2025

Meet MCP: The Cool New Way AI Remembers Stuff

IEEE CS PESU • Model Context Protocol (MCP) Workshop

A minimal MCP server that lists and sends Gmail messages via the Gmail API using Python.

Folder structure

c:\Abhay\PES\Workshop
├─ credentials.json        # OAuth client (Desktop) from Google Cloud Console
├─ token.json              # Generated after auth; stores user tokens
├─ requirements.txt        # Python dependencies for this repo
├─ scripts\
│  └─ auth.py              # One-time OAuth flow; creates/updates token.json
└─ src\
   ├─ gmail.py             # Gmail API helpers (auth, list emails, send email)
   └─ server.py            # MCP stdio server exposing list_emails and send_email

Prerequisites

• Python 3.9+ on Windows
• Gmail API enabled in a Google Cloud project
• OAuth client credentials (Application type: Desktop) downloaded as credentials.json

Setup (Windows, VS Code terminal)

1. Place credentials.json in the repo root:

2. Create a virtual environment and install deps:

   • py -3 -m venv .venv
   • .venv\Scripts\activate
   • pip install -r requirements.txt

3. Authenticate (opens browser; writes token.json to repo root):

   • python scripts\auth.py

4. Run the MCP server (stdio):

   • python src\server.py

Tools exposed

• list_emails
  • Input: { q?: string, maxResults?: number (1–50, default 5) }
  • Output: array of { id, threadId, snippet, from, subject, date }
• send_email
  • Input: { to: string, subject: string, text: string, cc?: string, bcc?: string }
  • Output: { status: "sent", id, threadId }

Troubleshooting

• 403 insufficientPermissions: enable Gmail API and re-auth with correct scopes.
• No refresh_token: delete token.json, ensure consent screen completes fully.
• Port conflicts on 3000: close the process using the port and retry auth.
• Stdio launch issues: ensure the client’s working directory is the repo root.