Skip to main content
Glama
peican996

weather-mcp

by peican996
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

Weather MCP Server + MCP Log Proxy

两个 MCP (Model Context Protocol) 工具，通过 stdio 协议与 MCP 客户端（如 Claude Code、Cline、VS Code 等）交互。

weather-mcp/
├── weather.py           # Weather MCP 服务器 — 查美国天气
├── main.py              # 占位入口
├── pyproject.toml        # 项目依赖
├── README.md
└── mcpLog/
    ├── mcp_log.py        # MCP Log Proxy — 透明代理记录请求/响应
    └── pyproject.toml    # proxy 依赖

1. Weather MCP Server

提供两个工具，数据来源 weather.gov API。

安装

# 安装 uv（如果还没有）
pip install uv

# 同步依赖
uv sync

工具

工具

参数

说明

get_alerts

state: string

查询美国州的天气警报（两字母州代码，如 CA）

get_forecast

latitude: float, longitude: float

查询经纬度处的天气预报

运行

uv run weather.py

Related MCP server: MCP Weather Server

2. MCP Log Proxy

透明 stdio 代理，记录 MCP 客户端与目标服务器之间的所有 JSON-RPC 流量。

Client ≤stdin/stdout≥ mcp_log.py ≤stdio≥ Target Server
                              │
                        logs/<timestamp>_<pid>.jsonl

安装

cd mcpLog
uv sync

命令行用法

python mcp_log.py <target_command> [target_args...] [--log-dir <dir>]

示例 — 代理 weather 服务器：

python mcp_log.py python ../weather.py --log-dir ./logs

日志格式

每行一个 JSON 对象，字段：

字段

说明

ts

UTC 时间戳

dir

META（元信息）/ REQUEST >>（请求）/ RESPONSE <<（响应）

msg

完整的 JSON-RPC 消息体

示例：

{"ts":"2026-06-13T15:17:31","dir":"REQUEST  >>","msg":{"method":"tools/call","params":{"name":"get_alerts","arguments":{"state":"CA"}},"jsonrpc":"2.0","id":5}}
{"ts":"2026-06-13T15:17:38","dir":"RESPONSE <<","msg":{"jsonrpc":"2.0","id":5,"result":{"content":[{"type":"text","text":"..."}]}}}

3. 配置 Cline（VS Code 插件）

编辑 %APPDATA%/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json：

直连 weather（不打日志）

{
  "mcpServers": {
    "weather": {
      "autoApprove": ["get_forecast", "get_alerts"],
      "disabled": false,
      "timeout": 60,
      "type": "stdio",
      "command": "uv",
      "args": [
        "--directory",
        "C:\\code\\python\\weather",
        "run",
        "weather.py"
      ]
    }
  }
}

通过 proxy 连接（记录所有请求日志）

{
  "mcpServers": {
    "weather": {
      "autoApprove": ["get_forecast", "get_alerts"],
      "disabled": false,
      "timeout": 60,
      "type": "stdio",
      "command": "C:\\code\\python\\weather\\mcpLog\\.venv\\Scripts\\python.exe",
      "args": [
        "C:\\code\\python\\weather\\mcpLog\\mcp_log.py",
        "C:\\code\\python\\weather\\.venv\\Scripts\\python.exe",
        "C:\\code\\python\\weather\\weather.py",
        "--log-dir",
        "C:\\code\\python\\weather\\mcpLog\\logs"
      ]
    }
  }
}

注意：路径里的 .venv\Scripts\python.exe 是 Windows 的 venv 路径，macOS/Linux 改为 .venv/bin/python。

配置后 reload Cline（或重启 VS Code）即可生效，日志输出到 mcpLog/logs/ 目录。

4. 配置 Claude Code

在项目根目录创建 .mcp.json：

{
  "mcpServers": {
    "weather": {
      "type": "stdio",
      "command": "python",
      "args": [
        "mcpLog/mcp_log.py",
        "python",
        "weather.py",
        "--log-dir",
        "mcpLog/logs"
      ]
    }
  }
}

依赖

  • Python ≥ 3.13

  • MCP SDK ≥ 1.27

  • httpx

  • anyio

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers