Summary
"Real math, not hallucinated numbers" is the entire argument, and it is the right one.
Ask a model for a portfolio's value at risk and it will produce a confident, plausible, wrong number — the arithmetic is the part language models are worst at, and risk figures are exactly where a plausible wrong answer does damage. Moving the computation to actual code against actual market data is the only way this belongs anywhere near a decision. Check what the free tier covers before depending on it; market data is the expensive input here.
What it is
A quantitative risk server that computes portfolio analytics against real market data rather than having a model estimate them.
What you get
- Value at risk and Monte Carlo simulation
- Stress testing and portfolio optimisation
- Option Greeks and correlation matrices
Requirements
Node, from npm. MIT licensed, with a free tier and a paid plan.
Setup effort
One command plus a key — npm install -g @quantrisk/mcp-server, then supply credentials