Skip to content

This project implements an MCP server that orchestrates the full workflow for agricultural modeling, from data gathering and preparation to simulation execution. It integrates with DSSAT and PYTHIA, automating input generation, managing data pipelines, and running simulations at scale.

License

Notifications You must be signed in to change notification settings

SCiO-systems/dssat-mcp

Repository files navigation

DSSAT MCP Server — Claude Desktop Integration

Connect Claude Desktop to a DSSAT MCP server to download inputs, run experiments, and collect outputs—directly from a chat.

Quick Start

1) Install Claude Desktop

  • Windows: Download and install the official Windows app, then sign in.

  • Ubuntu/Debian (community build): Use the community repo aaddrick/claude-desktop-debian to install a .deb/.AppImage, or build from source:

    git clone https://github.com/aaddrick/claude-desktop-debian.git
    cd claude-desktop-debian
    ./build.sh
    sudo dpkg -i ./claude-desktop_*.deb
    
    # If you encounter dependency issues:
    sudo apt --fix-broken install

Config paths:

  • Windows: %APPDATA%\Claude\claude_desktop_config.json

  • Linux: ~/.config/Claude/claude_desktop_config.json

    Tip: You can also open the config from Settings → Developer → Edit Config in Claude Desktop.

2) Add MCP configuration

Paste the JSON below into your config file. Replace only the token value; keep everything else identical. Tokens will be provided upon request.

{
  "mcpServers": {
    "dssat": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://dev.mcp.crops.qvantum.scio.services/mcp",
        "--allow-http",
        "--header",
        "Authorization: Bearer ${ACCESS_TOKEN}}"
      ],
      "env": {
        "ACCESS_TOKEN": "<REPLACE_WITH_YOUR_ACCESS_TOKEN>"
      }
    }
  }
}

If your config invokes npx, ensure Node.js and npm are installed.

3) Use it from Claude Desktop

  • Restart Claude Desktop (or reload the developer config).
  • Open a new chat. Claude Desktop will automatically discover the MCP server and list available tools.
  • Use the example prompt below to run a typical DSSAT workflow.

Tools (high-level description)

  • download_files_from_s3 — Download DSSAT input files (e.g., SOIL.SOL, *.WTH, *.MZX) from S3 into a local working folder and verify presence.
  • run_dssat_experiment — Run DSSAT using the specified experiment file (FileX) inside the working folder and capture logs/summary.
  • upload_and_collect_output_files — Archive the working folder, upload to S3, return a presigned URL for outputs, and perform cleanup.

Example Prompt (paste in Claude Desktop)

I want to run a DSSAT experiment. Check the list of tools to find the correct tool input format. First download and save to folder CUSTOM_EXPERIMENT the following files from S3: SOIL.SOL, UFGA8201.WTH, UFGA8201.MZX. Next execute the experiment with the experiment file UFGA8201.MZX. Then, upload the file to S3 in order to collect the output files.

About

This project implements an MCP server that orchestrates the full workflow for agricultural modeling, from data gathering and preparation to simulation execution. It integrates with DSSAT and PYTHIA, automating input generation, managing data pipelines, and running simulations at scale.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •