screen_rfq / screen_rfq_batch
        │
        ▼
┌─────────────────────────────────────┐
│  Node 1: parse_rfq                  │  ← PDF fetch + regex extraction
│  Node 2: check_flags                │  ← QPL/Berry/CMMC/ITAR/CAI/HAZMAT
│  Node 3: score_opportunity          │  ← 0–100 score across 5 dimensions
│  Node 4: lookup_grainger            │  ← Price + margin (skipped if disq'd)
└─────────────────────────────────────┘
        │
        ▼
   HITL Summary → Zapier → Gmail/Slack → You approve → DIBBS Fast Trak

cd fast_award_screener_mcp
pip install -r requirements.txt
python -m py_compile server.py && echo "✅ Syntax OK"

{
  "mcpServers": {
    "fast_award_screener": {
      "command": "python",
      "args": ["/path/to/fast_award_screener_mcp/server.py"],
      "env": {}
    }
  }
}

python server.py --transport streamable_http --port 8001
# Then expose via ngrok:
ngrok http 8001

fast_award_screener_mcp/
├── server.py           ← FastMCP server + 3 tools
├── state.py            ← Single RFQState TypedDict (consolidated)
├── requirements.txt
├── nodes/
│   ├── __init__.py
│   ├── parse_rfq.py        ← Node 1: PDF parsing
│   ├── check_flags.py      ← Node 2: Disqualifying clauses
│   ├── score_opportunity.py ← Node 3: Viability scoring
│   └── lookup_grainger.py  ← Node 4: Grainger price lookup
└── README.md