Weather MCP Server
Provides weather alerts and forecasts via the National Weather Service API, enabling Claude to answer weather-related queries.
README
Weather MCP Server for Claude
This repository implements a Model Context Protocol (MCP) server for Claude. It exposes weather-related tools via FastMCP over stdio, using the National Weather Service (NWS) API for data.
Project Structure
weather/
├── clients/
│ ├── __init__.py
│ └── nws_client.py
├── formatters/
│ ├── __init__.py
│ └── alert_formatter.py
├── services/
│ ├── __init__.py
│ └── weather_service.py
├── tools/
│ ├── __init__.py
│ └── weather_tools.py
├── models/
│ ├── __init__.py
│ └── responses.py
├── config.py
├── logging_config.py
├── main.py
├── pyproject.toml
└── README.md
Features
- MCP server for Claude using
FastMCP get_alerts(state)tool for active weather alertsget_forecast(latitude, longitude)tool for forecast data- Simple, maintainable service and client separation
Requirements
- Python 3.10 or higher
uvinstalled and available at the full path used by Claudemcpandhttpxdependencies installed in the Python environment
Running the Server Locally
From the repository root:
cd /Users/dinukakasun/Documents/MCP/weather
/Users/dinukakasun/.local/bin/uv --directory /Users/dinukakasun/Documents/MCP/weather run main.py
or using the virtual environment Python if preferred:
/Users/dinukakasun/Documents/MCP/weather/.venv/bin/python main.py
The server should start and expose MCP tools on stdio.
Claude Desktop MCP Configuration
Update ~/Library/Application Support/Claude/claude_desktop_config.json so Claude launches this server with the correct executable path and working directory.
Example configuration:
{
"mcpServers": {
"weather": {
"command": "/Users/dinukakasun/.local/bin/uv",
"args": [
"--directory",
"/Users/dinukakasun/Documents/MCP/weather",
"run",
"main.py"
]
}
}
}
Why absolute paths matter
Claude desktop may not inherit your interactive shell PATH. Using the absolute uv path avoids startup failures caused by missing commands.
Example Claude Questions
Use these example prompts to exercise the MCP server:
Get the current weather alerts for CA using get_alerts.Call get_forecast with latitude 37.7749 and longitude -122.4194.What are active alerts for NY?Provide a 5-period weather forecast for 40.7128, -74.0060.
Local Test Client
You can verify the MCP tools locally with a small Python client that starts the server and sends a request over stdio.
import asyncio
import json
import subprocess
async def run_request():
proc = await asyncio.create_subprocess_exec(
"/Users/dinukakasun/.local/bin/uv",
"--directory",
"/Users/dinukakasun/Documents/MCP/weather",
"run",
"main.py",
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
)
request = {
"jsonrpc": "2.0",
"id": 1,
"method": "initialize",
"params": {
"protocolVersion": "2024-11-05",
"capabilities": {},
"clientInfo": {"name": "local-test", "version": "1.0.0"}
}
}
stdout, stderr = await proc.communicate((json.dumps(request) + "\n").encode())
print(stdout.decode())
print(stderr.decode())
asyncio.run(run_request())
This confirms the server accepts MCP requests. For tool-specific tests, replace method with a valid tool name and pass params matching the tool signature.
Verifying the MCP Server
- Start the server manually.
- Confirm the process is running:
ps aux | grep "main.py" | grep -v grep
- Confirm Claude is configured to use the same command and working directory.
- If Claude still does not connect, use the absolute
uvpath inclaude_desktop_config.json.
Notes
- The MCP server uses
stdiotransport, not a network HTTP endpoint. - The entrypoint is
main.py, notweather.py. - The tools are registered in
tools/weather_tools.py.
Reference
- Model Context Protocol Python server docs: https://modelcontextprotocol.io/docs/develop/build-server#python
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。