simply-mcp 0.1.0

A modern Python framework for building MCP servers with multiple API styles

Simply-MCP-PY

A modern, Pythonic framework for building Model Context Protocol (MCP) servers with multiple API styles

Simply-MCP-PY is the Python implementation of simply-mcp-ts, bringing the same ease-of-use and flexibility to the Python ecosystem for building MCP servers.

Features

• Multiple API Styles - Choose the style that fits your workflow:

  • 🎨 Decorator API - Clean, declarative class-based approach
  • 🔧 Functional API - Programmatic server building with method chaining
  • 📝 Interface API - Pure type-annotated interfaces (coming soon)
  • 🤖 Builder API - AI-powered tool development (future)

• Multiple Transports - Run your server anywhere:

  • 📟 Stdio - Standard input/output (default)
  • 🌐 HTTP - RESTful HTTP server with session support
  • 📡 SSE - Server-Sent Events for real-time streaming

• Zero Configuration - Get started instantly:

  • Auto-detect API style
  • Automatic schema generation from type hints
  • Sensible defaults for everything
  • Optional configuration for advanced use cases

• Developer Experience:

  • 🔥 Hot reload with watch mode
  • 📦 Bundle to standalone executable
  • 🎯 Type-safe with full mypy support
  • 📚 Comprehensive documentation

• Production Ready:

  • 🔒 Rate limiting and authentication
  • ⚡ Progress reporting for long operations
  • 📊 Binary content support
  • 🛡️ Security best practices

Quick Start

Installation

Option 1: Try without installing (uvx)

# Install uvx
pip install uv

# Run directly without installing simply-mcp
uvx simply-mcp --version
uvx simply-mcp run server.py

Option 2: Install permanently (pip)

pip install simply-mcp

Your First Server (Decorator API)

# server.py
from simply_mcp import mcp_server, tool

@mcp_server(name="my-server", version="1.0.0")
class MyServer:
    @tool(description="Add two numbers")
    def add(self, a: int, b: int) -> int:
        """Add two numbers together."""
        return a + b

    @tool(description="Greet a user")
    def greet(self, name: str, formal: bool = False) -> str:
        """Generate a greeting."""
        if formal:
            return f"Good day, {name}."
        return f"Hey {name}!"

Run Your Server

# Run with stdio (default)
simply-mcp run server.py

# Run with HTTP on port 3000
simply-mcp run server.py --transport http --port 3000

# Run with auto-reload
simply-mcp run server.py --watch

Note: If using uvx, prefix commands with uvx: uvx simply-mcp run server.py. First run takes ~7-30 seconds to download packages, subsequent runs are near-instant.

API Styles

Decorator API (Recommended)

Clean, declarative class-based approach:

from simply_mcp import mcp_server, tool, prompt, resource

@mcp_server(name="my-server", version="1.0.0")
class MyServer:
    @tool(description="Calculate sum")
    def add(self, a: int, b: int) -> int:
        return a + b

    @prompt(description="Generate code review")
    def code_review(self, language: str) -> str:
        return f"Please review this {language} code..."

    @resource(uri="config://server", mime_type="application/json")
    def get_config(self) -> dict:
        return {"status": "running", "version": "1.0.0"}

Functional API

Programmatic server building:

from simply_mcp import BuildMCPServer

mcp = BuildMCPServer(name="my-server", version="1.0.0")

@mcp.add_tool(description="Add two numbers")
def add(a: int, b: int) -> int:
    return a + b

@mcp.add_prompt(description="Generate greeting")
def greet(name: str) -> str:
    return f"Hello, {name}!"

# Method chaining
mcp.configure(port=3000).run()

Configuration

Create a simplymcp.config.toml file:

[server]
name = "my-mcp-server"
version = "1.0.0"

[transport]
type = "http"  # or "stdio", "sse"
port = 3000

[logging]
level = "INFO"
format = "json"

[security]
enable_rate_limiting = true
rate_limit_per_minute = 60

Or use environment variables:

export SIMPLY_MCP_TRANSPORT=http
export SIMPLY_MCP_PORT=3000
export SIMPLY_MCP_LOG_LEVEL=DEBUG

Advanced Features

Progress Reporting

from simply_mcp import tool, Progress

@tool(description="Process large dataset")
async def process_data(data: list, progress: Progress) -> dict:
    total = len(data)
    for i, item in enumerate(data):
        await progress.update(
            percentage=(i / total) * 100,
            message=f"Processing item {i+1}/{total}"
        )
        # Process item...
    return {"processed": total}

Authentication

# simplymcp.config.toml
[security.auth]
enabled = true
type = "api_key"
api_keys = ["your-secret-key"]

Binary Content

@resource(uri="file://document.pdf", mime_type="application/pdf")
def get_document(self) -> bytes:
    with open("document.pdf", "rb") as f:
        return f.read()

CLI Commands

# Run a server
simply-mcp run server.py [--transport TYPE] [--port PORT] [--watch]

# Bundle to executable
simply-mcp bundle server.py --output dist/

# List available servers
simply-mcp list [--json]

# Configuration management
simply-mcp config init          # Create config file
simply-mcp config validate      # Validate config
simply-mcp config show          # Show current config

Examples

Check out the examples/ directory:

• simple_server.py - Minimal working example
• decorator_basic.py - Decorator API basics
• functional_api.py - Functional API usage
• http_server.py - HTTP transport example
• advanced_features.py - Progress, auth, binary content

Documentation

• Technical Specification - Detailed technical specs
• Architecture - System architecture
• Implementation Roadmap - Development plan
• API Reference - Complete API docs
• Migration Guide - Coming from simply-mcp-ts?

Comparison with simply-mcp-ts

Feature	simply-mcp-ts	simply-mcp-py
Decorator API	✅	✅
Functional API	✅	✅
Interface API	✅	🚧 (planned)
Builder API	✅	🚧 (future)
Stdio Transport	✅	✅
HTTP Transport	✅	✅
SSE Transport	✅	✅
Watch Mode	✅	✅
Bundling	✅	✅
Schema Validation	Zod	Pydantic
Type System	TypeScript	Python + mypy

Development Status

🚧 Currently in Alpha - Core features are implemented and functional, but the API may change. See ROADMAP.md for development progress.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Requirements

• Python 3.10 or higher
• Dependencies managed via pip

License

MIT License - see LICENSE file for details.

Related Projects

• simply-mcp-ts - TypeScript version
• Anthropic MCP SDK - Official Python MCP SDK
• Model Context Protocol - Protocol specification

Acknowledgments

• Built on top of the Anthropic MCP Python SDK
• Inspired by simply-mcp-ts
• Created by Clockwork Innovations

Support

• 📖 Documentation
• 🐛 Issue Tracker
• 💬 Discussions

---

Made with ❤️ by Clockwork Innovations