RTOFS MCP Server

MCP server for accessing NOAA RTOFS (Real-Time Ocean Forecast System) global ocean forecast data. Provides SST, salinity, ocean currents, and sea surface height from the operational HYCOM/ESPC model via the HYCOM THREDDS Data Server.

Features

• Global 1/12° ocean forecasts (~8 km resolution)
• 8-day forecast horizon, updated daily
• Temperature, salinity, currents at all depth levels
• Sea surface height (SSH) at hourly resolution
• Point queries, depth profiles, area grids, and vertical transects
• Pure httpx — no NetCDF/xarray dependencies needed

Installation

# Using uvx (recommended)
uvx rtofs-mcp

# Using pip
pip install rtofs-mcp

Tools

Tool	Description
rtofs_get_system_info	RTOFS overview — resolution, variables, coverage (no HTTP)
rtofs_list_datasets	List HYCOM THREDDS datasets with live availability check
rtofs_get_latest_time	Query latest forecast time from THREDDS
rtofs_get_surface_forecast	Surface time series at a point (SST, SSS, currents, SSH)
rtofs_get_profile_forecast	3D depth profile at a point (temp, salinity, currents vs depth)
rtofs_get_area_forecast	Surface forecast for a bounding box (parallel point queries)
rtofs_get_transect	Vertical transect between two points (parallel depth profiles)
rtofs_compare_with_observations	Change at a point over the 24 h before a time (default: now)

Example Queries

• "What is the current SST at The Battery, NYC?"
• "Show me a temperature depth profile at 35°N, 74°W in the Gulf Stream"
• "Get the SST forecast for the next 8 days at latitude 25, longitude -80"
• "Show a vertical transect of temperature from Miami to Bermuda"
• "What are the ocean currents at 40°N, 74°W?"

Data Source

HYCOM THREDDS Data Server (https://tds.hycom.org/thredds/)

ESPC-D-V02 datasets (operational HYCOM/RTOFS):

• FMRC_ESPC-D-V02_ssh — Sea Surface Height (2D, hourly)
• FMRC_ESPC-D-V02_t3z — Water Temperature (3D, 3-hourly)
• FMRC_ESPC-D-V02_s3z — Salinity (3D, 3-hourly)
• FMRC_ESPC-D-V02_uv3z — Ocean Currents u/v (3D, 3-hourly)

Data is queried via THREDDS NCSS (NetCDF Subset Service) which returns CSV for point queries — no NetCDF library required.

Configuration

Claude Desktop

{
  "mcpServers": {
    "rtofs": {
      "command": "uvx",
      "args": ["rtofs-mcp"]
    }
  }
}

Development

cd servers/rtofs-mcp
uv sync --group dev
uv run rtofs-mcp                                          # Start server
uv run pytest tests/ --ignore=tests/test_live.py -v       # Unit tests
uv run pytest tests/test_live.py -v                       # Integration tests

License

MIT