mcp-community 0.0.3

Easily run, deploy, and connect to MCP servers

Model Context Protocol Community

Goals:

• Community driven
• Implementation details are open-source
• Running, deploying, and connecting to MCP servers is dead simple
• Extremely clear documentation on:
  • How to build an MCP server (fill in the official docs' gaps)
  • How to run/deploy an MCP server
  • How to use an MCP client in your application
• Nice chat UI for using / testing MCP servers

Available Servers

You can find the list of available servers here

Getting Started

First, install the package:

pip install mcp-community

Next, try running the CalculatorMCP server:

from mcp_community import run_mcp
from mcp_community.servers import CalculatorMCP

run_mcp(CalculatorMCP)

By default, this will be running at http://0.0.0.0:8000/sse.

Now, connect to the server and run some calculations in a separate process:

import asyncio

from mcp_community import mcp_client


async def run() -> None:
    """Connect to an MCP server and use the calculator tools."""
    async with mcp_client("http://0.0.0.0:8000/sse") as session:
        # List available tools
        list_tools_result = await session.list_tools()
        print("Available Tools:")
        for tool in list_tools_result.tools:
            print(f"- {tool.name}: {tool.description}")

        # Add two numbers together
        args = {"a": 5, "b": 7}
        added = await session.call_tool("add", arguments=args)
        assert added.content[0].type == "text"
        print(f"Addition Result:  {args['a']} + {args['b']} = {added.content[0].text}")


asyncio.run(run())

Usage

MCP Community provides two methods:

• run_mcp: runs an MCP server as an SSE application (accepts Server or FastMCP)
• mcp_client: connects to the SSE endpoint for an MCP server

Hosted Servers

Coming soon...