Weather MCP Server
patilpkpratik/Chatgpt-MCP
Acesso a dados meteorológicos via MCP com integração ChatGPT Web
Ver no GitHubServidor MCP que expõe ferramentas de previsão do tempo integradas à API do Serviço Nacional de Meteorologia dos EUA. Fornece alertas por estado e previsões por coordenadas geográficas, com suporte para múltiplos modos de transporte (stdio, HTTP streamable, SSE).
Instalação
npx convez-ai patilpkpratik/Chatgpt-MCP1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Integrar previsão do tempo em aplicações Claude/ChatGPT
- Consultar alertas meteorológicos por estado
- Obter previsões detalhadas por latitude e longitude
Veredito

MCP server bem estruturado e documentado para integração de dados meteorológicos. Suporta múltiplos modos de transporte e exposição via ChatGPT Web, mas limitado ao ecossistema NWS (EUA).
README
A tradução para o português deste README ainda está na fila.
Weather MCP Server
MCP server that exposes weather tools backed by the US National Weather Service API.
Tools
get_alerts(state: str)get_forecast(latitude: float, longitude: float)
Requirements
- Python 3.10+
uv(recommended) orpip
Note: this project will not install on Python 3.9.
Install
Option A: uv (recommended)
uv sync
Option B: pip
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
Configuration
Copy and customize the environment template:
cp .env.example .env
Key variables:
MCP_TRANSPORT:stdio,streamable-http, orsseMCP_HOST: host bind for HTTP/SSE modesMCP_PORT: port bind for HTTP/SSE modesMCP_STREAMABLE_HTTP_PATH: streamable HTTP route (default/mcp)MCP_SSE_PATH: SSE route (default/sse)WEATHER_USER_AGENT: optional NWS user-agent override
Run Modes
1) Local stdio mode (for local MCP clients)
uv run weather.py --transport stdio
2) Streamable HTTP mode (for ChatGPT Web connectivity)
uv run weather.py --transport streamable-http
By default, the streamable HTTP endpoint is available at:
http://localhost:8000/mcp
You can also override host/port/path directly:
uv run weather.py --transport streamable-http --host 0.0.0.0 --port 8080 --streamable-http-path /mcp
Connect to ChatGPT Web
ChatGPT Web cannot directly spawn local stdio processes. You must provide a reachable MCP URL.
- Start the server in streamable HTTP mode.
- Expose localhost via a secure tunnel (for example, Cloudflare Tunnel or ngrok).
- Copy the HTTPS URL for your
/mcpendpoint. - In ChatGPT, open Settings -> Apps.
- Add a custom app using your MCP server URL.
- Confirm tools are discoverable and run test prompts.
Example tunnel commands:
# ngrok
ngrok http 8000
# cloudflared
cloudflared tunnel --url http://localhost:8000
If your tunnel URL is https://demo.example.ngrok-free.app, use:
https://demo.example.ngrok-free.app/mcp
Example test prompts:
- "Use weather tools to get active alerts for CA"
- "Use weather tools to get the forecast for 40.7128, -74.0060"
Troubleshooting
- If dependency install fails, ensure
uvis up to date. - If
uv syncfails with proxy authorization while downloading Python, configure your proxy credentials or use a preinstalled local Python 3.10+. - If
pip install -e .fails with "requires a different Python", switch to Python 3.10+ first. - If tool calls fail, check that
api.weather.govis reachable. - If ChatGPT cannot connect, verify your public HTTPS tunnel URL resolves to
/mcp. - If ChatGPT cannot connect, verify host/port/path values match your tunnel target.
- If there are no alerts for a state, the tool may correctly return no active alerts.
Security Notes
- This server is read-only with respect to weather data.
- Do not expose development servers publicly without access controls.
- Review app permissions in ChatGPT before enabling broad usage.