MCP Server Orchestration Project
An intelligent Model Context Protocol (MCP) orchestration system that demonstrates dynamic server coordination and AI-powered query routing. This project showcases how a primary MCP server can intelligently analyze natural language queries and dynamically call specialized MCP servers to provide comprehensive responses.
🧠 Architecture Overview
graph TD
A[User Query] --> B[People-Info Orchestrator]
B --> C[Claude AI Analysis]
C --> D{Query Category}
D -->|Developer Query| E[Start Developer-Info Server]
D -->|Designer Query| F[Start Designer-Info Server]
D -->|Other| G[Direct Response]
E --> H[Call get_developer_name]
F --> I[Call get_designer_name]
H --> J[Claude Enhancement]
I --> J
G --> J
J --> K[Natural Language Response]
K --> A
🎯 How It Works
- User asks a natural language question like "What is the developer's name?"
- People-info orchestrator receives the query
- Claude AI analyzes the query to determine intent and category
- Orchestrator dynamically starts the appropriate specialized server
- MCP client calls the right tool on the specialized server
- Raw data is enhanced by Claude AI into a natural response
- User receives a conversational, helpful answer
🏗️ Project Structure
MCP-Server/
├── people-info-server/ # 🧠 Main orchestrator (ADD THIS TO CLAUDE DESKTOP)
│ ├── index.js # Intelligent routing and MCP client logic
│ ├── package.json # Dependencies including MCP client SDK
│ ├── .env.template # Environment template
│ └── README.md # Detailed orchestrator documentation
├── developer-info-server/ # 👨💻 Developer data server
│ ├── index.js # Returns developer information
│ └── package.json # Simple MCP server
├── designer-info-server/ # 🎨 Designer data server
│ ├── index.js # Returns designer information
│ └── package.json # Simple MCP server
├── setup.sh # One-command setup script
├── QUICK_START.md # Super simple setup guide
└── README.md # This file
🚀 Quick Start
🚨 IMPORTANT FOR NEW USERS: If you're setting up this project on a different machine, you need to update 2 critical paths. See the Path Configuration Guide below.
Option 1: One-Command Setup
./setup.sh
Option 2: Manual Setup
cd people-info-server
npm install
cp .env.template .env
# Edit .env and add your ANTHROPIC_API_KEY
Add to Claude Desktop
Add only the orchestrator to your Claude Desktop MCP settings:
{
"mcpServers": {
"people-info": {
"command": "node",
"args": ["/Users/banik/Desktop/Projects2025/MCP-Server/people-info-server/index.js"],
"env": {
"ANTHROPIC_API_KEY": "your_anthropic_api_key_here"
}
}
}
}
🎮 Usage Examples
Natural Language Queries
Ask about developers:
<use_mcp_tool>
<server_name>people-info</server_name>
<tool_name>get_people_info</tool_name>
<arguments>
{
"query": "What is the developer's name?"
}
</arguments>
</use_mcp_tool>
Ask about designers:
<use_mcp_tool>
<server_name>people-info</server_name>
<tool_name>get_people_info</tool_name>
<arguments>
{
"query": "Tell me about the designer on the team"
}
</arguments>
</use_mcp_tool>
General questions:
<use_mcp_tool>
<server_name>people-info</server_name>
<tool_name>get_people_info</tool_name>
<arguments>
{
"query": "Who is the frontend developer?"
}
</arguments>
</use_mcp_tool>
🔧 Technical Features
🧠 AI-Powered Query Analysis
- Uses Claude 3.5 Sonnet to understand user intent
- Categorizes queries as developer, designer, or other
- Determines appropriate action (get name, get info, etc.)
🔄 Dynamic Server Orchestration
- Starts specialized MCP servers on-demand
- Maintains connections for efficiency
- Proper cleanup and resource management
🌐 Real MCP Protocol Communication
- Uses official MCP SDK for client-server communication
- Proper JSON-RPC protocol implementation
- Error handling and fallback mechanisms
💬 Natural Language Enhancement
- Raw server responses enhanced by Claude AI
- Conversational, context-aware answers
- Maintains query context throughout the flow
📊 Data Separation
Developer-Info Server
- Name: "Neick"
- Purpose: Stores and provides developer information
- Tool:
get_developer_name
Designer-Info Server
- Name: "Jesse"
- Purpose: Stores and provides designer information
- Tool:
get_designer_name
People-Info Orchestrator
- No data storage - pure orchestration logic
- AI analysis and MCP client functionality
- Response enhancement and natural language processing
🎯 Benefits of This Architecture
- Separation of Concerns: Each server has a single responsibility
- Scalability: Easy to add new specialized servers
- Natural Interface: Users can ask questions in plain English
- Dynamic Resource Usage: Servers only run when needed
- AI Enhancement: Raw data becomes conversational responses
- Real MCP Protocol: Demonstrates proper MCP client-server communication
🔍 Example Interaction Flow
User Query: "What is the developer's name?"
- Orchestrator receives query
- Claude analyzes:
{"category": "DEVELOPER", "action": "GET_NAME"} - Orchestrator starts developer-info server
- MCP client calls
get_developer_nametool - Server responds:
{"content": [{"type": "text", "text": "Neick"}]} - Claude enhances: "The developer's name is Neick."
- User receives natural response
🛠️ Development & Testing
Test Individual Servers
# Test developer server
cd developer-info-server && npm start
# Test designer server
cd designer-info-server && npm start
# Test orchestrator
cd people-info-server && npm start
Add New Specialized Servers
- Create new server directory (e.g.,
manager-info-server) - Implement MCP server with appropriate tools
- Update orchestrator's analysis logic to recognize new categories
- Add MCP client calls for the new server
🏆 Advanced MCP Concepts Demonstrated
- Multi-server orchestration
- Dynamic server lifecycle management
- MCP client-server communication
- AI-powered request routing
- Natural language query processing
- Response enhancement and formatting
🔧 Path Configuration for Sharing
If you're setting up this project on a different machine than the original developer, you need to update these paths:
1. Update Node.js Path in Code
Edit people-info-server/index.js line 84:
// Change this hardcoded path:
command: '/Users/banik/.nvm/versions/node/v22.16.0/bin/node',
// To use your system's Node.js:
command: 'node',
2. Update Project Path in Claude Desktop Settings
Replace the hardcoded path in your Claude Desktop MCP settings:
{
"mcpServers": {
"people-info": {
"command": "node",
"args": ["/YOUR/ACTUAL/PATH/TO/MCP-Server/people-info-server/index.js"],
"env": {
"ANTHROPIC_API_KEY": "your_anthropic_api_key_here"
}
}
}
}
Example paths:
- macOS:
"/Users/yourname/Desktop/MCP-Server/people-info-server/index.js" - Windows:
"C:\\Users\\yourname\\Desktop\\MCP-Server\\people-info-server\\index.js" - Linux:
"/home/yourname/Desktop/MCP-Server/people-info-server/index.js"
✅ What Doesn't Need Changes
- Developer-Info Server - Completely portable
- Designer-Info Server - Completely portable
- Internal server discovery - Uses relative paths automatically
📝 License
This project is open source and available under the MIT License.
This project showcases advanced MCP patterns and serves as a reference for building intelligent, AI-powered MCP orchestration systems. 🚀