ghh1125/xarray
0
xarray MCP Service
This deployment package wraps xarray with a FastMCP service and exposes practical tools for creating, selecting, reducing, aligning, and merging labeled arrays and datasets.
Available MCP Tools
health_checkcreate_dataarraycreate_datasetselect_dataarrayreduce_dataarrayalign_arraysmerge_datasetsadapter_overview
Detailed parameter and usage docs are in xarray/mcp_output/README_MCP.md.
Local stdio (Claude Desktop / CLI)
cd xarray/mcp_output
MCP_TRANSPORT=stdio python start_mcp.pyOr use local stdio entrypoint:
cd xarray/mcp_output/mcp_plugin
python main.pyHTTP transport (Docker / Spaces)
cd xarray/mcp_output
MCP_TRANSPORT=http MCP_PORT=8000 python start_mcp.pyMCP clients should connect to:
http://localhost:8000/mcp
For Hugging Face Spaces, Docker sets MCP_TRANSPORT=http and MCP_PORT=7860, so clients connect to:
https://<your-space-host>/mcp
Docker quick start
chmod +x run_docker.sh
./run_docker.shThe container listens on port 7860 (from port.json).
