Katabatic

The system of record for stablecoin reserve risk.

Cornell AI Hackathon 2026 · Programmable Capital Track

---

Getting Your Machine Ready

Prerequisites

Install these before anything else:

Tool	Version	Install
Python	3.11, 3.12, or 3.13 (NOT 3.14)	brew install python@3.12
Node.js	18+	brew install node
Git	any	brew install git
GitHub CLI	any	brew install gh then gh auth login

1. Clone the repo

git clone https://github.com/AI-HackathonNYC/AI-Hackathon-Will-change-name-.git
cd AI-Hackathon-Will-change-name-

2. Run the setup script

chmod +x setup.sh
./setup.sh

This will:

• Create a Python virtual environment in backend/venv/
• Install all Python dependencies (FastAPI, anthropic, google-genai, networkx, etc.)
• Copy .env.example to .env (you'll fill in keys next)
• Install all Node dependencies for the frontend
• Run a health check to verify the backend starts

3. Add your API keys

Edit backend/.env and fill in your keys:

# Required — get these BEFORE the hackathon
ANTHROPIC_API_KEY=sk-ant-...        # https://console.anthropic.com
GEMINI_API_KEY=...                   # https://aistudio.google.com/apikey
ETHERSCAN_API_KEY=...               # https://etherscan.io/myapikey (free tier)
PINATA_API_KEY=...                  # https://app.pinata.cloud/developers (free tier)
PINATA_SECRET_API_KEY=...           # Same Pinata dashboard

# No key needed: NOAA (User-Agent only), Open-Meteo (public), FDIC BankFind (public), Nominatim (User-Agent only)

DO NOT commit this file. It's in .gitignore.

4. Verify everything works

Start the backend:

cd backend
source venv/bin/activate
uvicorn main:app --reload

You should see:

INFO:     Uvicorn running on http://127.0.0.1:8000

Visit http://localhost:8000/health — you should get {"data":{"status":"ok"},...}

Start the MCP server (optional, for AI agent integration):

cd backend
source venv/bin/activate
python mcp_server.py

Start the frontend (in a separate terminal):

cd frontend
npm run dev

You should see:

VITE v6.x.x  ready in Xms
➜  Local:   http://localhost:5173/

Start the slides (if needed, separate terminal):

cd slides-app
npm run dev

5. You're ready

If both health check and frontend dev server start without errors, you're good to go.

---

Project Structure

├── backend/          ← FastAPI API + scoring engine (Python)
├── frontend/         ← React dashboard (TypeScript)
├── slides-app/       ← Pitch deck presentation (React)
├── data/             ← Fixture fallbacks, extracted data, cache
├── scripts/          ← One-off data processing scripts
├── setup.sh          ← One-command dev setup
├── CLAUDE.md         ← Full project spec
└── TASKS.md          ← Master task list

Live MCP Endpoint

The Helicity MCP server is deployed on Blaxel and accessible as a hosted function:

https://run.blaxel.ai/{workspace}/functions/katabatic-mcp/mcp

This endpoint uses the streamable-http transport and exposes all 5 MCP tools: get_stress_scores, get_stablecoin_detail, project_scenario, get_active_alerts, get_score_history.

AI agents can connect directly to this URL using any MCP-compatible client.

---

Collaborators

• Adi Prathapa
• Aritro Ganguly
• Connor Mark
• Krish Jana
• Suchit Basineni
• Vikram Davey