Skip to main content

1. Connect

Open https://quantx.limex.pro, sign in, and follow the on-site instructions to attach QuantLab MCP to your IDE. Reload MCP servers when you are done — the QuantLab tools should appear in the client.

2. Ask for something concrete

In chat, try:
Equal-weight backtest on AAPL, MSFT, GOOGL from 2023-01-01 to 2025-01-01 — no ML.
Or open Example prompts for longer, more realistic requests (stress tests, charts, ML, full optimization sweeps).

3. Reuse earlier results

Every finished step returns a link to its result, with a unique id. These URLs illustrate the format only; use the actual output_url from your own successful job:
Paste your returned link back into the chat to continue from it instead of downloading the data again. A data_extractor_ link feeds portfolio construction or feature engineering, and a portfolio_optimization_ link feeds a backtest, stress test, or risk report.

4. Let the session remember them for you

Rather than keeping links yourself, you can let QuantLab record them. Ask it to enable the session registry — at the start of a longer piece of work it will usually offer — and it writes every successful result to .quantspace/quantspace-session.json in your project, then reads that file whenever one step feeds the next. The file is plain JSON, local to that project, and works the same in Cursor, Claude Code and any other client that can read and write files. Add .quantspace/ to .gitignore. Without it, nothing breaks — you just supply the links yourself.