Idea into code you can run
Describe the trade in plain language. The app writes Python and runs it on your computer, so you can read the code it executed.
Quant research agent · macOS, Windows, and Linux
Test the idea before you trust it
Cursor, Claude Code, and Codex can write the script. Atabet Quant Intelligence is a quant research agent, built by practitioners, that runs on your computer. Describe a trading idea. It loads market data, writes Python, and backtests it locally — costs, delay, and whether the execution could have happened. You get a test you can rerun, including whether an idea that may not hold up should be rejected. Download the installer, then subscribe to activate it.
What the quant work actually is
A general coding assistant stops at a script. This app carries the idea through data, a local Python run, and a backtest that is allowed to fail.
Atabet is also a quant research desk. Technical analysis, factor studies, event-driven backtests, and market-regime detection are built in as modules on the same workflow. While you work, you can speak directly with a quantitative researcher, so the next step rests on methods that already exist.
A backtest, not a chat
The run includes the checks quants argue about: costs, delay, and whether that fill could have happened. Tables and charts are saved in the open project.
Know when to reject it
The useful result is a test you can run again. If the idea does not survive the checks, that is the answer. This is research software, not investment advice.
Product Demo
Watch one pass: a trading idea in plain language, market data, a local Python backtest, and a report saved in the project.
Demo video coming soon
Add public/videos/product-demo.mp4 (H.264) to enable playback.
How It Works
Three steps from a trading idea to code, a backtest, and a file you can open again.
Step 1
State the idea
Describe the hypothesis and how you would trade it. You do not start from a blank file.
Step 2
Run the test locally
The agent loads market data and runs Python on your computer, with quant modules that flag unrealistic execution.
Step 3
Verdict on disk
The project folder holds the HTML report, table, or chart — the test itself, not only a chat reply.