Yug771/MCP-Server

DocAssistant: MCP Server & Client

This project demonstrates a simple client-server application for document management using the fastmcp library. The server provides a set of tools and resources for interacting with text files, and the client script shows how to use them. Communication between the client and server is handled over standard input/output (stdio), making it a lightweight and efficient solution for local inter-process communication.

Project Structure

MCP-Server-And-Client/
├── documents/            # Stores all text documents
│   └── notes.txt         # An example document
├── MCP-client.py         # Example script to interact with the server
├── MCP-server.py         # The backend server logic
├── app.py                # Cloud-ready Streamlit app (no MCP dependencies)
├── requirements.txt      # Python dependencies for cloud deployment
├── requirements_local.txt # Full dependencies for local development
└── README.md             # This file

Features

The DocAssistant server exposes several fastmcp endpoints:

Resources

• docs://list: Lists all available .txt documents in the documents directory.
• docs://{filename}: Retrieves the full content of a specified document.

Tools

• append_to_doc: Appends a string of content to the end of a specified file.
  • Parameters: filename (str), content (str)
• search_in_doc: Searches for a keyword within a document and returns the line numbers and content of all matches.
  • Parameters: filename (str), keyword (str)

Prompts

• append_prompt: Generates a standardized message for appending content, intended to guide a user or another AI.
• search_prompt: Creates a standardized message for searching content.

Setup & Installation

1. Prerequisites

• Python 3.8+
• An activated virtual environment (recommended)

2. Install Dependencies

For cloud deployment (Streamlit Cloud):

pip install -r requirements.txt

For local development with full MCP functionality:

# First, save the original requirements if needed
cp requirements.txt requirements_local.txt
# Then install from the local requirements
pip install fastmcp mcp

3. Running the Application

Option 1: Command Line Interface (Local Only)

The client and server run in a single, coordinated process. To start the application, simply run the client script from your terminal:

python MCP-client.py

The client will automatically start the server as a subprocess and begin interacting with it. You will see output from both the client and server in your terminal, demonstrating the various features of the DocAssistant.

Option 2: Web Interface (Works on Cloud)

For a more user-friendly experience, you can use the Streamlit web interface:

streamlit run app.py

This will open a beautiful web application in your browser where you can:

• List all available documents
• View document contents with syntax highlighting
• Append new content to documents
• Search for keywords within documents
• Download documents

The app.py version works on Streamlit Cloud as it doesn't require MCP subprocess communication.

Cloud Deployment

Deploying to Streamlit Cloud

1. Push your code to GitHub
2. Go to share.streamlit.io
3. Connect your GitHub repository
4. Select app.py as the main file
5. Deploy!

Note: The cloud version (app.py) implements document operations directly without MCP client-server architecture, making it compatible with cloud environments.

How It Works

Local Version (MCP-client.py + MCP-server.py)

The MCP-client.py script initializes a StdioTransport to manage communication with the MCP-server.py script. It then uses the fastmcp client to:

1. Ping the server to ensure it's running.
2. List all available tools, resources, and prompts.
3. Call the append_to_doc tool to add a new line to documents/notes.txt.
4. Call the search_in_doc tool to find occurrences of the word "Hello".
5. Read the entire content of documents/notes.txt using its resource URI.
6. Generate a prompt message using the append_prompt template.

Cloud Version (app.py)

The cloud version directly implements file operations without requiring a separate server process, making it suitable for deployment on platforms like Streamlit Cloud where subprocess communication is restricted.