mcp_gmail_calender

Langchain MCP Agent

This project demonstrates a Langchain agent integrated with multiple MCP (Multi-tool Coordination Protocol) servers, including a custom Gmail MCP server, a Google Calendar MCP server, and other utility tools like math and search.

Setup and Running Locally

To run this project on your local system using uv, follow these steps:

1. Clone the repository:

   git clone https://github.com/harshdabhi/mcp_gmail_calender.git
   cd mcp_gmail_calender
   
   

2. Install uv:

   If you don't have uv installed, you can install it using pip:

   pip install uv
   

   Or, if you prefer a standalone executable, follow the instructions on the official uv documentation.

3. Install Python Dependencies:

   Ensure you have all the necessary Python packages installed. From the project root, run:

   uv pip install -r requirements.txt
   

4. Set up Environment Variables (.env file):

   Create a .env file in the root of your project and add your GROQ_API_KEY:

   GROQ_API_KEY=your_groq_api_key
   

   Replace your_groq_api_key with your actual Groq API key.

5. Configure MCP Servers:

   The project uses a mcp_config/mcp_servers_file.json to configure the MCP servers. Ensure this file is correctly set up with the paths to your local MCP server implementations.

6. Gmail MCP Server Setup:

   For detailed instructions on how to enable and authenticate the Gmail MCP server, refer to its documentation: https://glama.ai/mcp/servers/@GongRzhe/Gmail-MCP-Server

   A common step involves moving your gcp-oauth.keys.json to ~/.gmail-mcp/ and running an npx command for authentication:

   mkdir -p ~/.gmail-mcp/
   mv <path_to_your_gcp-oauth.keys.json> ~/.gmail-mcp/
   npx @gongrzhe/server-gmail-autoauth-mcp auth
   

   Note: You need Node.js and npm installed for npx to work. If you encounter npx: command not found, please install Node.js and npm (e.g., using brew install node on macOS).

7. Google Calendar MCP Server Setup:

   For detailed instructions on how to enable and use the Google Calendar MCP server, refer to its documentation: https://glama.ai/mcp/servers/@cablate/mcp-google-calendar

   Similar to Gmail, you might need to set up authentication and ensure GOOGLE_CALENDAR_ID, GOOGLE_TIME_ZONE, and GOOGLE_CREDENTIALS_PATH are correctly configured, preferably via environment variables in your .env file.

8. Run the Langchain Agent:

   Once all dependencies are installed and configurations are in place, you can run the Langchain agent:

   python server/langchain_server
   

   This will start an interactive chat in your terminal. Type your queries, and the agent will respond using the configured MCP tools. Type exit or quit to end the session.

LangSmith Tracing

This project is configured to use LangSmith for tracing agent interactions. To enable tracing, ensure you have the following environment variables set in your .env file:

LANGCHAIN_TRACING_V2=true
LANGCHAIN_API_KEY=your_langsmith_api_key
LANGCHAIN_PROJECT=your_project_name  # Optional: Replace with your desired project name

Replace your_langsmith_api_key with your actual LangSmith API key.

After setting these variables, run the agent as described in step 8 of the "Setup and Running Locally" section. Your traces will automatically appear in your LangSmith account.

Video Demonstration