MCP

This MCP server is a Python-based Modular Command Platform setup that enables you to expose simple arithmetic tools and dynamic resources as conversational actions. In the quickstart example, you create a FastMCP instance and register a tool function (e.g., add) that performs a calculation, as well as a dynamic resource (e.g., greeting) that forms a personalized response. The server can be extended to integrate with Claude (via the Pro plan) for AI-assisted dialogue, and can be wired to Zoho CRM later to provide CRM-aware conversational capabilities. To run the server, execute the main script (server.py) with Python: python server.py. Once running, the MCP framework will listen for requests and route them to the registered tools/resources, enabling clients to call add(2, 3) to get 5 or get_greeting(name) to receive a personalized greeting.

Prerequisites:

• Python 3.x installed
• Node.js installed (for any backend tooling, optional for this Python-focused setup)
• Visual Studio Code (optional for development)
• Claude Pro Plan (if you plan to integrate Claude AI in the workflow)

Step 1: Create and prepare the project

• Create a new project folder and open it in your editor.
• Ensure you have a Python environment available (virtualenv or conda is optional but recommended).

Step 2: Set up the MCP project with uv (Python package manager)

• Install uv: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
• Initialize the MCP project (example name: mcp-server-demo): uv init mcp-server-demo cd mcp-server-demo

Step 3: Add MCP to dependencies and install the server script

• Add MCP CLI support: uv add "mcp[cli]"
• Install the MCP server script referencing your main Python file (example: server.py): uv run mcp install main.py

Step 4: Create the MCP server script

• Create a file named server.py with your MCP definitions (example provided in the README): from mcp.server.fastmcp import FastMCP

  mcp = FastMCP("Demo")

  @mcp.tool() def add(a: int, b: int) -> int: """Add two numbers""" return a + b

  @mcp.resource("greeting://{name}") def get_greeting(name: str) -> str: """Get a personalized greeting""" return f"Hello, {name}!"

  if name == "main": mcp.run()

Step 5: Run the server

• Start the MCP server: python server.py

Notes:

• The above steps assume a Python-based MCP setup. If you switch to a different runtime (e.g., Node.js), adjust the commands accordingly.

Tips and tips:

• Keep the server script modular: separate tools (functions) from resources (data-fetchers) for easier extension.
• If integrating Claude AI, ensure API access and relevant configuration are set up in your environment as per Claude’s Pro plan requirements.
• Use environment variables to store sensitive config (API keys, endpoints) rather than hard-coding them.
• When testing locally, enable verbose logging or debugging in your mcp server to diagnose tool invocation and routing issues.
• For production deployments, consider containerizing the server (e.g., Docker) and wiring health checks to keep the MCP service reliable.