MCP Server Demo

MCP Server Demo with Streamlit and Ollama

This project is a demo of a Model Context Protocol (MCP) server in Python, integrated with a web chat interface built in Streamlit and a local language model using Ollama. It allows users to manage users in a PostgreSQL database (add, delete, modify, and query) both from the chat and from external MCP tools.

Characteristics

• Web chat with Streamlit.
• Local LLM using Ollama (Llama 3 or other compatible model).
• MCP server with tools for user CRUD.
• PostgreSQL database for storing users.
• Flexible integration : You can use just the chat, just the MCP server, or both.

Requirements

• Python 3.9+
• PostgreSQL running and accessible
• Ollama installed and running (for local LLM)
• Virtual environment (recommended)
• Dependencies in the requirements.txt file

Facility

1. Clone the repository:
   git clone https://github.com/tu-usuario/mcp-server-demo.git cd mcp-server-demo
2. Create and activate a virtual environment:
   python -m venv .venv source .venv/bin/activate # En Linux/Mac .venv\\Scripts\\activate # En Windows
3. Install the dependencies:
   pip install -r requirements.txt
4. Configure your PostgreSQL database and configure the connection details in server.py .
5. (Optional) Create a .env file if you use API keys.

Use

1. Levanta Ollama and the LLM model

2. Open the Streamlit chat

Open your browser at http://localhost:8501 .

3. (Optional) Start the MCP server

either

Chat Features

• Add user:
  Write: agregar usuario Juan juan@mail.com clave123
• Delete user:
  Type: eliminar usuario juan@mail.com
• Modify user:
  Type: modificar usuario juan@mail.com nombre=Juan Perez contraseña=nuevaClave
• Consult users:
  Write: consultar usuarios or consultar usuario juan@mail.com
• Free conversation:
  If the message is not an action, the local LLM responds.

Credits

• Streamlit
• Ollama
• psycopg2
• Model Context Protocol (MCP)

Contributions and suggestions are welcome!