MCP server brings institutional options screening to Claude Desktop
Connerlambden built Helium MCP, a free remote MCP server that exposes per-symbol ML options pricing models to Claude and Cursor, enabling traders to screen for ratio spread opportunities in 30 seconds without institutional infrastructure.
Score breakdown
Developers and traders can now query institutional-grade ML options pricing models directly from Claude or Cursor with zero setup cost, enabling rapid screening for structural mispricings and ratio spread opportunities that previously required expensive Bloomberg infrastructure and custom models.
- 01Helium MCP is a free, no-signup remote MCP server that exposes per-symbol ML options pricing models trained on each ticker's historical data, returning predicted fair value, probability of expiry in the money, and data freshness for any contract
- 02The server integrates into Claude Desktop or Cursor via a single `mcp.json` configuration entry pointing to https://heliumtrades.com/mcp
- 03Connerlambden demonstrates screening AAPL May 15 options and identifies a 1x2 put ratio spread opportunity by comparing Helium's model fair values (which assign ~1% probability to finishing below $195) against market prices showing rich premiums for deep OTM puts
Helium MCP is a free remote MCP server built by Connerlambden that exposes per-symbol ML options pricing models to LLMs like Claude and Cursor. The server is a thin wrapper over a machine-learning model trained on each ticker's historical options data, returning three key outputs for any contract: predicted_price (the model's fair value), prob_itm (probability of expiring in the money), and options_data_date (snapshot freshness). Installation requires a single addition to `mcp.json` with the command `npx -y mcp-remote https://heliumtrades.com/mcp`.
The same workflow extends to calendar spread arbitrage, diagonal spreads, and volatility compression screening.
The article demonstrates a practical use case: screening for 1x2 put ratio spreads, a classic options strategy where traders sell one near-the-money put and buy two out-of-the-money puts to express a view that downside skew is overpriced. Connerlambden queries AAPL May 15 expiry options across multiple strikes and observes that the model assigns ~1% probability to finishing below $195, yet the market prices deep OTM puts significantly higher—a textbook skew mispricing. The same workflow extends to calendar spread arbitrage, diagonal spreads, and volatility compression screening. Helium MCP also exposes a `get_top_trading_strategies` endpoint that ranks long-vol vs short-vol candidates with explicit bull/bear narratives. By making ML option pricing accessible via MCP, the tool shifts what was a two-decade institutional moat into a zero-marginal-cost query from inside an LLM.
Key facts
- 01Helium MCP is a free, no-signup remote MCP server that exposes per-symbol ML options pricing models trained on each ticker's historical data, returning predicted fair value, probability of expiry in the money, and data freshness for any contract
- 02The server integrates into Claude Desktop or Cursor via a single `mcp.json` configuration entry pointing to https://heliumtrades.com/mcp
- 03Connerlambden demonstrates screening AAPL May 15 options and identifies a 1x2 put ratio spread opportunity by comparing Helium's model fair values (which assign ~1% probability to finishing below $195) against market prices showing rich premiums for deep OTM puts
- 04The per-symbol ML models make different pricing calls than generic Black-Scholes fits, sometimes wildly different, because they are trained on each ticker's own historical options data
- 05Helium MCP also exposes a `get_top_trading_strategies` endpoint that returns daily-ranked long-vol vs short-vol screens with explicit bull/bear cases and skew analysis