MCP Python Server—API Wrapper

This project creates an MCP server in Python that exposes a tool for querying an external API. It is compatible with Claude Desktop or ChatGPT Desktop that support the Model Context Protocol (MCP).

✨ Features

Exposing a tool via MCP
HTTP query to an external API
Direct integration with Claude/Desktop via claude.json

🚀 Requirements

Python 3.9+
mcp[cli] (installable via pip or uv)
Claude or ChatGPT Desktop (with MCP support)

📁 Project structure

.
├── servidores/profile.py          # Servidor MCP con herramientas para interactuar con mi backend del curriculum.
├── server.py                     # Servidor MCP con herramienta "consultar_api".
├── .env                          # Variables opcionales para auth/API.
├── claude.json                   # Config. MCP para integrarlo directamente.
└── README.md                     # Este documento.

⚙️ Installation

With pip

pip install "mcp[cli]"

With UV (recommended)

uv init mcp-api-server
cd mcp-api-server
uv add "mcp[cli]"

Installation of the MCP

mcp install mi_script.py

Installation with .env

mcp install mi_script.py -f .env

Installation of dependencies

pip install -r requirements.txt

Environment variables

Create a .env file in the project root to define optional environment variables:

# .env
API_KEY=mi_api_key
API_URL=https://miapi.com/consulta

👷 Quick Start (Quickstart)

Create the server server.py

from mcp.server.fastmcp import FastMCP
import httpx

mcp = FastMCP("API Wrapper")

@mcp.tool(description="Consulta una API externa")
async def consultar_api(param: str) -> str:
    """Consulta una API externa con un parámetro y devuelve la respuesta."""
    async with httpx.AsyncClient() as client:
        r = await client.get(f"https://miapi.com/consulta?param={param}")
        return r.text

Run locally in dev mode

mcp dev server.py

Run in production mode

mcp run server.py

Or with uv:

uv run --with mcp[cli] mcp run server.py

🚀 Integration with Claude/Desktop

Locate claude.json in the Claude/Desktop configuration folder:

On Windows: %APPDATA%\Claude\claude.json
On Linux/macOS: ~/.claude/claude.json

Example:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/codigo/backend-curso-inkor/proyectos_memes"
      ]
    },
    "Demo": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "mcp",
        "run",
        "C:\\codigo\\backend-curso-inkor\\MCP\\server.py"
      ]
    }
  }
}

🤖 Use within Claude/Desktop

You can ask the model:

Use the consult_api tool with the "ping" parameter

And the model will use your MCP server to make a real-time HTTP call.

🎁 Bonus: Tools Extension

@mcp.tool()
async def traducir(texto: str, lang: str) -> str:
    return f"Traducido: {texto} → {lang}"

🔍 Resources

Official MCP documentation: https://docs.mcp.run/
Python SDK repository: https://github.com/modelcontextprotocol/mcp

✅ Made with love and httpx 🚀

MCP Python Server — API Wrapper

Integrations