G4 MCP Server Implementations

A multi-language implementation of a minimal, extensible Model Context Protocol (MCP) server.

This repository contains multiple implementations of the same MCP architecture in different languages. Each implementation follows the same design principles:

• JSON-RPC based MCP protocol
• Tool definitions separated from execution logic
• Reflection-based tool discovery
• SSE-compatible streaming endpoint
• Designed for VS Code Agent Mode and automation platforms like n8n

---

Table of Contents

• Overview
• Architecture Philosophy
• Available Implementations
  • Python (Flask)
• MCP Capabilities
• When to Use Each Implementation
• Folder Structure
• Contribution Guidelines
• Roadmap

---

Overview

This repository provides reference implementations of an MCP (Model Context Protocol) server.

MCP enables:

• LLM tool calling
• Agent-based execution
• Structured tool schemas
• Tool discovery
• Streaming-capable transport

Each language implementation exposes:

• initialize
• tools/list
• tools/call
• SSE endpoint (optional but recommended)

All implementations follow the same tool design rules:

1. Tool metadata defined in JSON (or schema file)
2. Tool logic implemented as a function/method
3. Tool name must match definition name
4. All tools return structured JSON objects

---

Architecture Philosophy

The architecture is intentionally:

• Minimal
• Explicit
• Reflection-based
• Stateless
• Easy to extend
• Agent-first

Core layers:

Definitions  →  Cache  →  Domain  →  Repository  →  Transport

Where:

• Definitions = JSON schemas describing tools
• Cache = In-memory storage of tool metadata
• Domain = Dispatcher and orchestrator
• Repository = Tool implementations
• Transport = HTTP + JSON-RPC + SSE

---

Available Implementations

Python (Flask)

Location:

src/python/

Full documentation:

👉 Python Implementation README

Features:

• Flask-based MCP server
• SSE endpoint
• JSON-RPC POST endpoint
• Reflection-based tool loading
• Docker-ready
• IDE (VSCode, PyCharm, Visual Studio, etc.) Agent compatible
• n8n Streamable HTTP compatible

---

MCP Capabilities

All implementations aim to support:

Capability	Description
JSON-RPC	Standard MCP method dispatch
Tool Discovery	tools/list
Tool Execution	tools/call
Structured Output	structuredContent support
Streaming	SSE endpoint
Agent Mode Compatible	VS Code Agent
Automation Compatible	n8n Tools Node

---

When to Use Each Implementation

Python (Flask)

Best for:

• Rapid prototyping
• Lightweight deployments
• Educational reference
• Docker-based environments
• Edge deployments
• Quick MCP testing

---

Folder Structure

From repository root:

/README.md
/src
   /python
      README.md
      app.py
      domain.py
      cache_manager.py
      tools_repository.py
      definitions/

Each language implementation lives under src/<language>/.

---

Contribution Guidelines

If adding a new language implementation:

1. Follow the same MCP method contract
2. Keep tool definition separation
3. Support JSON-RPC POST
4. Optionally support SSE
5. Include a README.md inside the language folder
6. Add a link to it in this root README

---

Roadmap

Planned future implementations:

• .NET
• Node.js
• FastAPI
• Go
• Rust
• Cloud-native Kubernetes-ready variant

---

Design Goal

This repository is intended to serve as:

• A reference MCP architecture
• A clean starting point for agent-enabled systems
• A minimal, understandable tool-first MCP server
• A foundation for enterprise-grade automation