usda-mcp/
├── clients/
│   ├── nass_client.py       # NASS QuickStats API wrapper
│   └── ams_client.py        # AMS Market News API wrapper
├── server/
│   ├── main.py              # MCP server — 4 tools
│   ├── security.py          # OWASP MCP Top 10 mitigations
│   └── tools/
│       └── __init__.py
├── tests/
│   └── qa_log.csv           # Q&A evaluation log
├── demo/
│   └── app.py               # Demo interface (planned)
├── logs/
│   └── usda_nass_server.log # Auto-generated server log
├── .env                     # API keys (never committed)
├── .env.example             # Template for API keys
├── requirements.txt         # Pinned dependencies
├── SECURITY.md              # OWASP MCP Top 10 documentation
└── README.md

git clone https://github.com/RolakeOkans/usda-mcp.git
cd usda-mcp

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

pip install -r requirements.txt

NASS_API_KEY=your_nass_key_here
AMS_API_KEY=your_ams_key_here

{
  "mcpServers": {
    "usda-nass": {
      "command": "/path/to/your/python",
      "args": ["/path/to/usda-mcp/server/main.py"],
      "env": {
        "NASS_API_KEY": "your_key_here",
        "AMS_API_KEY": "your_key_here"
      }
    }
  }
}

User asks a plain English question
        ↓
AI model reads the question and decides which tool to call
        ↓
MCP server validates and sanitizes the input
        ↓
NASS or AMS API returns the raw data
        ↓
Server scans response for security issues
        ↓
AI model interprets the data and returns a plain English answer
        ↓
All tool calls logged to logs/usda_nass_server.log