Overview
What is Model Context Protocol (MCP)?
Model Context Protocol (MCP) is a custom MCP server that gives any AI application access to a toolset for an Artificial Virtual Assistant (AVA). It integrates with Gmail via Google OAuth, allowing the assistant to read and send emails. This server is designed for developers building AI agents with email capabilities.
How to use Model Context Protocol (MCP)?
Clone the repo, install uv, then test the server with uv run mcp dev mcp-server-example.py. Add the server configuration to an AI app (e.g., Claude Desktop or Cursor) using the provided JSON config. Customize the assistant’s behavior by editing prompts/ava.md, set up a .env file with your Google OAuth credentials, and authorize Gmail access by running uv run oauth.py.
Key features of Model Context Protocol (MCP)
- Integrates with Gmail via Google OAuth
- Customizable assistant personality and preferences
- Testable in dev mode with
uv run mcp dev - Compatible with Claude Desktop and Cursor
- Secure credential and token management
- Simple setup with uv package manager
Use cases of Model Context Protocol (MCP)
- Building an AI virtual assistant with email capabilities
- Automating email tasks like reading and sending messages
- Creating agentic workflows that run on desktop AI apps
- Personalizing assistant behavior for specific user preferences
FAQ from Model Context Protocol (MCP)
What are the runtime dependencies for Model Context Protocol?
You need Python, uv (the package manager), and Google OAuth credentials for Gmail API access.
How do I authorize Gmail access for the server?
Run uv run oauth.py – it checks for an existing token, initiates Google OAuth authentication in your browser, and saves the token automatically.
Where are credentials and tokens stored?
Credentials go in .config/ava-agent/credentials.json and the OAuth token is stored in .config/ava-agent/token.json. Both paths are configured in your .env file.
Can I customize the assistant’s behavior?
Yes. Edit the prompts/ava.md file to update communication preferences, specific instructions, and any other guidelines for the assistant.
