Skip to content

This is the repository for the AI Forward Developer Take-Home Challenge for CALMHSA. In this I built out an MCP Server that integrates into Claude Desktop and designed a few AI agents to utilize said MCP server

Notifications You must be signed in to change notification settings

keidson299/AIForwardDeveloperChallenge

Repository files navigation

MCP Server Setup Guide

Overview

This guide explains how to set up and run the MCP server, execute tests, and integrate with Claude Desktop.

Prerequisites

  • Python 3.10 or higher
  • fastmcp package installed (pip install fastmcp)
  • Claude Desktop application

Project Structure

├── main.py # MCP server implementation
├── test_mcp_tools.py # Test suite
├── config/
│ └── claude_desktop_config.json
├── logs/ # Log files directory
│ ├── mcpServer.log
│ └── work_log.json
└── tasks.json # Task management storage

This Project was built and tested in VSCode, and the instructions reflect that. You may have to modify the instructions if you are using a different IDE

Getting Started

1. Setup Required Directories

First, create the necessary directories and files by running these commands in a powershell terminal:

mkdir logs
echo "[]" > tasks.json
echo "[]" > logs/work_log.json

2. Install FastMCP

In a powershell terminal, run the command:

pip install fastmcp

3. Starting the MCP Server

In a PowerShell terminal run the command:

python main.py

You should see a FastMCP startup text, and no further output as the server communicates through stdin/stdout.

4. Running Tests

Open a second PowerShell terminal and run:

python test_mcp_tools.py

Expected Output:

Starting MCP server tests...

Testing analyze_file...
File analysis result: {
"lineCount": 5,
"hasTodos": true,
"hasFunctions": true,
"hasComments": true
}
✅ analyze_file test passed

Testing log_work...
...

Claude Desktop Integration

1. Configure Claude Desktop

  1. Open Claude Desktop
  2. Go to Settings > Developer
  3. Enable Developer Mode
  4. Update the MCP configuration:
{
   "mcpServers": {
      "software-support": {
         "command": "python",
         "args": ["path/to/server/file/main.py"],
      }
   }
}

2. Testing the Integration

  1. In Claude Desktop, create a new chat
  2. Send a test message to use one of the MCP tools:
Can you analyze the file test_mcp_tools.py using the analyze_file tool?

Available MCP Tools

  • analyze_file: Analyzes code files for metrics
  • log_work: Records work activities
  • add_task: Creates new tasks
  • list_tasks: Shows all tasks
  • complete_task: Marks tasks as completed

Troubleshooting

Common Issues

  1. Server Not Starting

    • Check Python version:
    python --version
    • Verify fastmcp installation:
    pip list | findstr fastmcp
  2. Claude Desktop Connection Failed

    • Verify the path in claude_desktop_config.json matches your project location
    • Ensure Developer Mode is enabled
    • Check Claude Desktop logs for connection errors
  3. Tests Failing

    • Make sure the MCP server is running in a separate terminal
    • Check if tasks.json and log files exist
    • Review logs/mcpServer.log for errors

Logs

Check the following logs for troubleshooting:

  • Server logs: logs/mcpServer.log
  • Work logs: logs/work_log.json

For additional support, review the error messages in the terminal output or check the Claude Desktop developer console.

About

This is the repository for the AI Forward Developer Take-Home Challenge for CALMHSA. In this I built out an MCP Server that integrates into Claude Desktop and designed a few AI agents to utilize said MCP server

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages