Multi-Server-MCP

Multi MCP Agent Project

A modular, extensible agent system using the MCP (Model Context Protocol) to connect multiple tool servers (math, text utilities, code execution, and more) to a FastAPI backend and a simple web frontend. The agent leverages Azure OpenAI for natural language understanding and can be extended with additional MCP servers, both local and remote.

---

Features

• FastAPI backend for handling user queries and serving the frontend
• LangChain agent powered by Azure OpenAI
• MCP protocol for tool server integration
• Local MCP servers for math, text utilities, and code execution
• Support for external MCP servers via .vscode/mcp.json (see notes below)
• Simple HTML frontend for user interaction

---

Project Structure

Multi MCP/
│
├── main.py                  # FastAPI backend
├── client.py                # Agent logic and MCP client config
├── math_server.py           # Math operations MCP server
├── text_utilities_server.py # Text summarization, translation, keyword extraction MCP server
├── code_execution_server.py # Code execution MCP server
├── frontend.html            # Web UI
├── requirements.txt         # Python dependencies
├── .env                     # Environment variables (not committed)
└── .vscode/
    └── mcp.json             # External MCP server configs (see below)

---

Setup Instructions

1. Clone the Repository

git clone <https://github.com/balogunadebola/Multi-Server-MCP.git>
cd Multi MCP

2. Install Python Dependencies

pip install -r requirements.txt

3. Configure Environment Variables

Create a .env file in the project root with your Azure OpenAI credentials:

AZURE_OPENAI_API_KEY=your-azure-openai-key
AZURE_OPENAI_ENDPOINT=your-azure-endpoint
AZURE_OPENAI_VERSION=2023-05-15

4. (Optional) Configure External MCP Servers

Edit .vscode/mcp.json to add or modify external MCP servers. Note: Dynamic runners like uvx/uvenv are deprecated and not supported. For external servers, use direct HTTP endpoints or run the server locally and point to it with command and args.

---

Running the Application

Start the FastAPI server:

python main.py

This will:

• Start the backend on http://127.0.0.1:8000/
• Launch all configured MCP servers (math, text utilities, code execution, and any in .vscode/mcp.json that use a valid local runner)

Open frontend.html in your browser to use the web interface.

---

Available MCP Servers & Tools

Math Server (math_server.py)

• Addition, subtraction, multiplication, division, square

Text Utilities Server (text_utilities_server.py)

• Text summarization
• Translation (placeholder)
• Keyword extraction

Code Execution Server (code_execution_server.py)

• Run Python code
• Simulate JavaScript execution (placeholder)
• Run shell commands

External Servers (via .vscode/mcp.json)

• You can add external MCP servers by specifying their HTTP endpoint or a local script. Dynamic remote runners are not supported.

---

Extending the System

• Add new local MCP servers by creating a Python script and registering tools with the FastMCP decorator pattern.
• Add external servers by editing .vscode/mcp.json (HTTP endpoints or local scripts only).
• Update client.py and main.py if you add new local servers.

---

Security Notes

• The code execution server can run arbitrary Python and shell code. Use with caution and never expose to untrusted users.
• Keep your .env file secure and never commit it to version control.

---

Credits

• LangChain
• Azure OpenAI
• MCP Protocol

---

License

MIT License