Your First Hour

This page gets you from a fresh install to a real backtest with a real number on it. Not a demo, not a sandbox — the same loop you will use every day after this.

Budget an hour. Most of it is two install steps that have nothing to do with QuantIDE: an AI to drive it, and Python for it to run.

Step 1: Install QuantIDE

Download the build for your platform and open it. QuantIDE is a desktop application — it runs on your machine, keeps your files on your disk, and does not need a browser. It is one plan, billed monthly.

Sign in when it asks. The left rail holds the places you will work: Code, Stacks and Chat.

Step 2: Connect an AI

The assistant in every stack is driven by an AI engine run by the app on your machine: Claude Code from Anthropic, or Codex from OpenAI. Install whichever you prefer from its maker's page and sign in to it.

• Claude Code needs a paid Claude plan (Pro or Max).
• Codex signs in with your ChatGPT account.

Then open a chat in QuantIDE. If it cannot find either one, it shows a card for each with a link to get it, and a "Check again" button for once you have. If you just installed one and it still is not found, restart QuantIDE so it can see it.

Step 3: Install Python

QuantIDE ships no Python of its own. This is deliberate, and it is worth understanding on day one: the app is not a sealed runtime with a fixed list of blessed libraries. It runs whatever is on your machine, which means you are never waiting for us to support the tool you want.

The cost of that choice is this step. The research stack needs Python 3 with a few libraries. Check what you have from any terminal:

python3 --version

If that prints a version of 3.9 or higher, you are fine. If it errors, install Python 3 from python.org first.

Then install the libraries the research stack expects:

pip install vectorbt quantstats

Step 4: Open the Quant Research stack

On the Stacks tab you will find Quant Research already there. It ships with the app because an empty shelf teaches nobody what a stack is.

Click it. The workspace opens in two halves — a chat on the right, and the Canvas on the left, currently empty.

Step 5: Ask it something real

Describe a strategy idea in plain English. Not a prompt, not pseudocode — the way you would say it to another person:

Test a 10/50 moving average crossover on AAPL from 2020 to 2024.

Watch what happens, because the sequence is the product:

• 1.It restates your idea as testable rules — universe, signal, entry, exit, sizing — and names every assumption it had to invent to fill your gaps.
• 2.It writes one self-contained Python script into your working folder.
• 3.It runs that script and the result appears on the Canvas: an equity curve, a trade list, a metrics panel.
• 4.It reads the result back to you like a skeptic, not a salesman.

The script it wrote is a real file in a real folder. Click "Open in Code" in the workspace header and you are looking at it, in an editor, with a terminal. You can run it yourself. You can change a number and run it again. Nothing is hidden behind the chat.

Step 6: Read the answer properly

You now have numbers. The temptation is to look at the return. Resist it — a return figure on its own is the least informative number on the screen.

Look in this order: Sharpe and max drawdown first, then the trade count, then whether the gains came from the whole period or from three good weeks. Fewer than about thirty trades proves nothing at all.

What you just learned

You ran the loop the entire product is built around: idea, rules, script, backtest, verdict. Every other page in these docs is either about doing one of those steps better, or about the machinery that made it possible.

Two pieces of machinery made it possible: the AI that did the typing, which the next page covers in full, and the Stack that told it what method to follow.