mytool3
A lightweight MCP server that bridges Claude AI with local Python execution, enabling personalized greetings and demonstrating local tool integration.
README
🤖 MCP Server mytool3
A lightweight Model Context Protocol (MCP) server that provides a bridge between Claude AI and your local Python environment.
🌟 Overview
- mytool3 is a proof-of-concept MCP server built to demonstrate the seamless integration of local Python execution with Claude Desktop.
- It allows Claude to "step out" of its sandbox and interact with your machine via a standard tool interface.
📁 Project Structure
mcp_gen/
├── .venv/
├── mcpgen/ # generator logic
├── ├── __init__.py
| ├── cli.py
├ ├── generator.py
│ ├── mytool3/ # MCP server
│ │ ├── __init__.py
│ │ ├── server.py
│ │ └── mcp.json
├── .gitignore
├── README.md
└── pyproject.toml
⚒️ Features Standard Discovery:
- Automatically announces its tools to Claude via the list_tools protocol.
- Synchronous Execution: Uses stdio transport for high-speed, local communication.
- Adaptive Integration: Fully compatible with the Claude Desktop interface.
⚒️ Installation & Setup
- Environment Initialization Standardize the local environment using a virtual environment to isolate the MCP dependencies.
# Clone the repository
git clone https://github.com/reory/mytool3.git
cd mytool3
Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate # Automation for Mac/Linux
**OR: .venv\Scripts\activate (Windows)**
Upgrade core tooling
pip install --upgrade pip setuptools
- Dependency Resolution This project leverages the fastmcp SDK for streamlined protocol handling.
# Install required MCP packages
pip install fastmcp mcp
Verify installation
python -c "import fastmcp; print(f'FastMCP version: {fastmcp.version}')"
- Service Registration (Claude Desktop) To hook the server into the Claude Desktop runtime, inject the absolute paths into your local configuration. Command-line shortcut (PowerShell): PowerShell
Open the specific MSIX config path for the Windows Store version
- notepad "$env:LOCALAPPDATA\Packages\Claude_pzs8sxrjxfjjc\LocalCache\Roaming\Claude\claude_desktop_config.json" JSON Payload:
JSON { "mcpServers": { "mytool3": { "command": "C:\Absolute\Path\To\.venv\Scripts\python.exe", "args": ["C:\Absolute\Path\To\mcpgen\mytool3\server.py"] } } }
🧪 Development Workflow
Manual Protocol Test Before testing in the Claude UI, ensure the server initiates the stdio transport without Python exceptions:
# Set unbuffered mode to prevent pipe-hangs
export PYTHONUNBUFFERED=1
python mcpgen/mytool3/server.py
- Hot-Reloading (Optional) If iterating on tool logic, use the fastmcp dev-tools for immediate feedback:
fastmcp dev mcpgen/mytool3/server.py
🔧 Tools Available
| Tool Name | Parameters | Description |
|---|---|---|
hello |
name (string) |
Returns a personalized greeting from the local server. |
⚠️ Troubleshooting
- Silent Failures: If the "Running" badge is blue but no tools appear, ensure your server.py is returning CallToolResult and not CallToolRequest.
- The Ghost Sandbox: If changes to the code aren't reflecting, use Task Manager to End Task on all Claude processes to force a config reload.
- Import Errors: Ensure you run pip install mcp specifically for the Python version defined in your command path.
- It specifically calls out the 2026 UI features like the "Connectors" menu, which didn't exist in older versions of the app.Path Accuracy: It preserves that long, annoying "Packages" path we found—this is the #1 thing that trips people up.
- Code Logic: It warns about the Result vs Request trap we just solved.
🛣️ Roadmap Features
-
[ ] Persistent Memory (SQLite Integration) Goal: Give Claude a "long-term memory" that persists across different chat sessions.
-
[ ] File System Sentinel (Local I/O) Goal: Allow Claude to safely inspect and summarize local directory structures.
-
[ ] Web Research Bridge (Playwright/Scraper) Goal: Enable Claude to fetch real-time data from sites without official APIs (like documentation pages).
📁 Notes
- This was challenging to build as FastMcp is a new library in python and the documentation is new and still evolving. I had to cross reference documents from a few different sources. FastMcp
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。