Modular MCP Server with Python Tools

This project is a test/demo implementation to understand how to build a modular MCP (Model Context Protocol) server in Python, integrate custom tools, and use prompts to interact with language models.

Purpose

• Explore Python async programming and tool registration with MCP.

• Practice building tools that can be invoked via chat input (e.g., note creation, searching, running commands).

• Experiment with prompt design and using local LLMs (via Ollama) to parse natural language into structured commands.

• Understand how to detect user intents from free text and map them to specific tools.

• Learn how to handle tool invocation responses asynchronously and display results.

Related MCP server: mcp-server-llmling

Features

• Tool examples:

  • Note creation and search with SQLite backend.

  • Weather fetching.

  • Mathematical calculation.

  • Time queries.

  • Running shell commands safely.

  • File operations.

• Intent detection via keyword matching and prompt parsing.

• Ollama local LLM integration for structured data extraction.

• Modular design allowing easy addition of new tools.

Usage

• Start the MCP server (e.g., python simple_modular_server.py).

• Run the client script to chat and interact with tools.

• Use natural language commands like:

  • "make a note. title is shopping list. category groceries."

  • "search note groceries"

  • "run command 'pwd'"

  • "what's the weather in London?"

Notes

• This project is purely for learning and testing.

• It is not production-ready.

• Designed to help understand MCP architecture, Python tooling, and prompt engineering with local LLMs.

Requirements

• Python 3.8+

• Ollama installed and running locally with your preferred models (e.g., phi, llama3.2:3b).

• SQLite (built-in with Python).

• Dependencies listed in requirements.txt (if created).

License

This is an educational project with no license.

Feel free to experiment and extend!