Model Context Protocol (MCP) Server

An implementation of Anthropic's Model Context Protocol following the official specification from https://modelcontextprotocol.io

This server implements the MCP v1.0 specification and provides:

Core Features

• Document context processing with citation support
• Code context processing with language-specific handling
• Data context processing with structured results
• Health check endpoint with capabilities reporting
• Error handling and validation
• Support for context-specific options

Features

• REST API implementation of MCP
• Support for text, code, and data context types
• Proper error handling and metadata support
• FastAPI-based server
• Aligned with official MCP specification

Requirements

• Python 3.8+
• FastAPI
• Uvicorn

Installation

pip install fastapi uvicorn

Running the Server

python mcp_server.py

Client Configuration

To configure this MCP server in your client application, add the following to your configuration:

{
    "mcpServers": {
        "mcp_server": {
            "command": "python",
            "args": [
                "mcp_server.py"
            ],
            "env": {
                "MCP_PORT": "8000",
                "MCP_HOST": "0.0.0.0"
            }
        }
    }
}

Key Configuration Options:

• command: The Python interpreter command
• args: Arguments to pass to the command
  • Server filename (e.g., "mcp_server.py")
• env: Environment variables for server configuration
  • MCP_PORT: Port to run the server on (default: 8000)
  • MCP_HOST: Host to bind the server to (default: 0.0.0.0)

Claude Desktop Integration Guide

Prerequisites

1. Claude Desktop app installed (version 1.2.0 or later)
2. Python 3.8+ installed
3. MCP server running locally
4. Proper client configuration as shown above

Step-by-Step Setup

1. Install Dependencies

   pip install -r requirements.txt
   

2. Start the MCP Server

   python mcp_server.py
   

   The server will start on http://localhost:8000

3. Configure Claude Desktop

   • Open Claude Desktop app
   • Go to Settings > Integrations > MCP Servers
   • Click "Add New Server"
   • Enter the following details:
     • Name: Local MCP Server
     • URL: http://localhost:8000
     • Description: Local MCP server for testing
   • Click "Save"

4. Test the Connection

   • In the MCP Servers list, click the "Test Connection" button next to your server
   • You should see a green checkmark if the connection is successful

5. Using MCP in Conversations

   • Start a new conversation in Claude Desktop
   • Use the @mcp prefix to route queries to your MCP server
   • Example queries:

     @mcp Analyze this code snippet: [your code here]
     @mcp What can you tell me about this document?
     @mcp Process this data: [your data here]
     

Troubleshooting

• Connection Failed: Ensure the MCP server is running and accessible
• Timeout Errors: Check your network settings and firewall rules
• Invalid Responses: Verify the server logs for any errors

Example Conversation Flow

User: @mcp What can you tell me about this code?
def factorial(n):
    if n == 0:
        return 1
    return n * factorial(n-1)

Claude: [MCP Response] Processed code query: This appears to be a recursive implementation of the factorial function...

API Endpoints

• POST /context
  • Request:

    {
      "query": "your query",
      "context_type": "text|code|data",
      "context": {},
      "metadata": {}
    }
    

  • Response:

    {
      "response": "Processed response",
      "context_type": "text|code|data",
      "metadata": {},
      "error": null
    }
    

Future Plans

• Add authentication and security
• Implement data source integrations
• Add support for streaming responses
• Implement context caching
• Add support for context versioning