Built a free tool that computes independent ML fair values for every listed options contract

I've been working on an ML model that prices options independently from the market, then compares to actual market prices to surface mispricings. Wanted to share since I think some of you might find it useful.

**What it does:**

For any ticker, the model computes its own theoretical value (helium_theo) for every listed contract, alongside standard Greeks. It also generates:

– "Should I buy/sell" probability signals for each contract

– AI-ranked strategies sorted by expected value

– Backtested win rates per strategy type (e.g., short vol calls on AAPL: 61% win rate, avg +$8.40/trade over 39 historical trades)

– Full historical chains with ML pricing baked in

**Example — NVDA right now:**

The AI generates 5 probability-weighted scenarios:

– 38%: Mean-reversion, +0 to +7%, holds 175-185

– 25%: +10 to +25% on AI/data-center headlines

– 20%: -5 to -12%, puts stay bid

– 10%: -20 to -35%, export/H200 shock

– 7%: +25 to +40%, strong risk-on + upside guidance

Each scenario has specific falsifiability criteria so you know what would invalidate the thesis.

**How to access it:**

It's available for free via the MCP protocol — you add one line to your AI assistant config (Cursor, Claude, etc.) and can query it in natural language. No API key needed for the free tier.

GitHub: https://github.com/connerlambden/helium-mcp

Docs: https://heliumtrades.com/mcp-page/

Curious what you all think. The mispricings between helium_theo and market prices are where I've found the most interesting setups.