Meteo

A lightweight Model Context Protocol server that exposes weather forecasts using the Open-Meteo Météo-France API.

This MCP server is designed to provide weather forecast data to AI assistants through a simple, read-only interface. It uses Open-Meteo's Météo-France forecast endpoint, which provides access to high-resolution French weather models without requiring an API key.

What It Does

This MCP server uses the Open-Meteo Météo-France API, with support for Météo-France forecast models including AROME France, AROME France HD, AROME 15-minute variants, and ARPEGE. For French locations, the server can use Météo-France AROME France HD, a high-resolution model with approximately 1.5 km spatial resolution. Depending on configuration, Open-Meteo can also use a "best match" mode combining AROME and ARPEGE models.

This server lets an MCP-compatible assistant retrieve weather forecasts for a given location.

It is especially useful for:

• local weather forecasts in France
• daily and hourly weather summaries
• temperature, rain, wind and cloud cover analysis
• assistant workflows that need structured weather context
• lightweight self-hosted weather integrations

Data Source

This project uses the Open-Meteo Météo-France API:

https://open-meteo.com/en/docs/meteofrance-api

Open-Meteo provides access to several weather models and forecast variables through a free HTTP API.

The Météo-France endpoint is particularly relevant for forecasts in France and nearby regions.

Security Model

This MCP server is read-only.

It does not:

• write data
• expose personal information
• require user authentication
• store user data
• execute system actions
• require API secrets

Because it only exposes public weather forecast data, it can be safely deployed as a public read-only MCP endpoint if the server itself is properly isolated.

Architecture

MCP Client
  -> HTTPS endpoint
  -> MCP server
  -> Open-Meteo Météo-France API

Example with Caddy:

Internet
  -> Caddy HTTPS
  -> local MCP server
  -> Open-Meteo API

Requirements

• Python 3.11+
• MCP-compatible client
• Internet access from the server to Open-Meteo
• Optional: Caddy or another reverse proxy for public HTTPS exposure

Installation

Clone the repository:

git clone https://github.com/jducerf/mcp-meteo.git
cd mcp-meteo

Create a virtual environment:

python3 -m venv .venv
source .venv/bin/activate

Install dependencies:

pip install -r requirements.txt

If the project does not include a requirements.txt yet, install the dependencies used by server.py.

Run Locally

Example:

python server.py

Depending on your MCP runtime, you may also run it through an MCP proxy or service wrapper.

Example Deployment With systemd

Example environment file:

PORT=8012
MCP_COMMAND=python
MCP_ARGUMENT=server.py

Example service restart:

sudo systemctl restart [email protected]
sudo systemctl status [email protected] --no-pager -l

Example Caddy Configuration

meteo.example.com {
    reverse_proxy 127.0.0.1:8012
}

For a public read-only weather MCP, no OAuth layer is required as long as the server only exposes public forecast data and does not provide write or system capabilities.

Test Endpoint

curl -i https://meteo.example.com/mcp \
  -H 'Accept: application/json, text/event-stream'

Expected behavior:

• the server responds through the MCP transport
• no OAuth challenge is required
• no 401 Unauthorized for public read-only access

Why Open-Meteo?

Open-Meteo is a practical choice because it provides:

• free access
• no API key requirement
• structured JSON responses
• multiple forecast models
• high-resolution forecast data
• simple HTTP endpoints
• good compatibility with self-hosted tools and AI assistants

License

MIT