shreyashprashu/mcp

MCP Tools (Flask + OpenAI Bridge)

This project provides a Minimal Control Protocol (MCP) server exposing simple utility tools, along with a bridge to OpenAI’s API that enables automatic tool-calling inside chat completions.

It includes multiple components:

• MCP Server (Flask) – A JSON-RPC API exposing tools over HTTP.
• MCP Server (stdio) – An alternative stdio-based MCP server for local clients.
• Client – Example async client demonstrating how to connect and call MCP tools.
• OpenAI Bridge – Connects MCP tools to OpenAI’s chat.completions endpoint, letting GPT automatically discover and call tools.

🚀 Features

• Implements the MCP protocol over both HTTP (Flask) and stdio.

• Exposes a catalog of tools:

  • echo – Returns back a string.
  • add_numbers – Sums a list of numbers.
  • now – Returns the current datetime (with optional timezone).
  • word_count – Counts words and characters in text.

• JSON-RPC compliant endpoints (listTools, callTool, initialize).

• CORS-friendly Flask server for web integrations.

• Full OpenAI integration: tools are dynamically mapped into OpenAI’s tool schema so GPT models can invoke them seamlessly.

• Example client to test MCP servers directly.

📂 Project Structure

mcp/
├── server.py              # stdio-based MCP server (async, model-driven)
├── mcp_plain_flask_app.py # HTTP MCP server with Flask
├── client.py              # Example MCP client (stdio)
└── mcp_openai_bridge.py   # Bridge MCP → OpenAI tool-calling

⚙️ Installation

Requirements: Python 3.10+

pip install flask requests openai python-dateutil pytz mcp

🖥️ Running the MCP Servers

1. Flask MCP Server (HTTP)

Start the HTTP server (serves /mcp on port 8080):

python mcp_plain_flask_app.py

Health check:

curl http://localhost:8080/health

List available tools:

curl -X POST http://localhost:8080/mcp \
  -H "Content-Type: application/json" \
  -d '{"method":"listTools","id":1,"jsonrpc":"2.0"}'

Call a tool:

curl -X POST http://localhost:8080/mcp \
  -H "Content-Type: application/json" \
  -d '{"method":"callTool","id":2,"jsonrpc":"2.0","params":{"toolName":"echo","arguments":{"text":"Hello from MCP"}}}'

2. MCP Server (stdio)

Run the stdio server:

python server.py

3. Example Client

Run the sample client (connects to stdio server):

python client.py

4. OpenAI Bridge

Start the bridge (serves /chat on port 8090):

export OPENAI_API_KEY=sk-...
python mcp_openai_bridge.py

Query via chat:

curl -s -X POST http://localhost:8090/chat \
  -H 'Content-Type: application/json' \
  -d '{"prompt":"What time is it in Asia/Kolkata and how many words in \"hello brave new world\"?"}'

📝 Example Response

{
  "answer": "The current time in Asia/Kolkata is 2025-09-04T19:10:10 (UTC+05:30). The phrase \"hello brave new world\" contains 4 words."
}