skewt-mcp

An MCP (Model Context Protocol) server that generates Skew-T Log-P diagrams from weather model data for any coordinates. Designed for pilots, paragliders, and weather nerds who use AI assistants.

No API key needed — uses free Open-Meteo data (which blends HRRR, GFS, and other models).

What You Can Ask Your AI

Once installed, just talk naturally:

"Show me the skew-t for Ellenville, NY at 2pm"

"What does the sounding look like for 37.7749, -122.4194 tomorrow at 18Z?"

"Pull a skew-t for the Torrey Pines launch site — is it thermic today?"

"Get me the sounding data for my home airport, I want to check the inversion layer"

The MCP tool description tells the AI when and how to call it:

Generate a Skew-T Log-P diagram for given coordinates and time. Returns a PNG image and key thermodynamic indices (CAPE, CIN, LCL, LFC, EL, PWAT).

Example Output

For Ellenville Flight Park (41.67, -74.40) at 18Z:

CAPE: 0.0 J/kg
CIN: -0.1 J/kg
LCL: 991.1 hPa (75 m AGL)
LFC: 958.4 hPa
EL: 949.6 hPa
PWAT: 26.0 mm
0-6 km Shear: 44.8 kt

Plus a full Skew-T diagram with temperature (red), dewpoint (green), parcel path (dashed), wind barbs, CAPE/CIN shading, dry/moist adiabats, and mixing ratio lines.

Features

• Skew-T diagrams with temperature, dewpoint, parcel path, wind barbs, CAPE/CIN shading, dry/moist adiabats, and mixing ratio lines
• Computed indices: CAPE, CIN, LCL, LFC, EL, precipitable water, 0-6 km bulk shear
• Data source: Open-Meteo API (free, no API key) — uses best available model blend including HRRR for CONUS
• GFS fallback: If the default model fails, automatically falls back to GFS
• MCP transport: stdio — works with Claude Desktop, Claude Code, and any MCP client

Installation

pip install skewt-mcp

Or from source:

git clone https://github.com/keithmgould/skew-t_mcp
cd skew-t_mcp
pip install -e .

MCP Client Configuration

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "skewt": {
      "command": "python",
      "args": ["-m", "skewt_mcp"]
    }
  }
}

Claude Code

claude mcp add skewt -- python -m skewt_mcp

Tools

get_skewt

Generate a Skew-T Log-P diagram for given coordinates and time.

Parameters:

Parameter	Type	Required	Default	Description
latitude	float	✅	—	Latitude in decimal degrees
longitude	float	✅	—	Longitude in decimal degrees
forecast_hour	int	❌	current hour	Hour in UTC (0-23)
date	string	❌	today	Date in YYYY-MM-DD format
model	string	❌	"hrrr"	"hrrr" or "gfs"

Returns: PNG image (base64) + text summary with CAPE, CIN, LCL, LFC, EL, PWAT, and wind shear.

get_sounding_data

Get raw sounding data as JSON for programmatic use. Same parameters as get_skewt.

Returns: JSON with pressure levels, temperatures, dewpoints, wind speed/direction, and all computed indices. Useful for custom analysis or when you don't need the image.

How It Works

1. Fetches pressure-level data (1000–200 hPa, 19 levels) from the Open-Meteo API
2. Computes dewpoints from temperature + relative humidity (Magnus formula)
3. Calculates thermodynamic indices using MetPy
4. Renders the Skew-T diagram with matplotlib + MetPy's SkewT class
5. Returns the image as base64 PNG + a text summary

Development

git clone https://github.com/keithmgould/skew-t_mcp
cd skew-t_mcp
pip install -e ".[dev]"
pytest

Dependencies

• MetPy — meteorological calculations + Skew-T plotting
• matplotlib — rendering
• httpx — async HTTP client
• mcp — MCP Python SDK

License

MIT