RocketShip
StartEvaluation

Multi-agent LLM equity screener

Point it at 500 stocks.
See what it picks.

A deterministic score ranks the universe, four analyst agents argue over the survivors, and a judge decides. You get every memo, every verdict and the portfolio it builds — then you can check whether any of it actually predicted anything.

Run the scannerOr build your own screen →
Every argument, in fullFour analysts, one judgeBull, bear, regime and value each write a memo citing the actual metrics. The judge reads all four plus the underlying data, then returns a verdict, a confidence and what would change its mind.
A real portfolio8–12 positions, weightedA convex optimiser sets sizes under sector and concentration caps, with the position limits and every promotion or drop recorded so you can see exactly why a name made it.
And a way to check it19,051 stock-datesEvery stage is scored against realised forward returns, so you can tell which part of the pipeline earns its place — and rebuild the screen yourself if it does not.

The pipeline under test

  1. 01

    Screen

    A deterministic score ranks the universe on momentum, volume, quality and sector alignment. No LLM involved.

  2. 02

    Debate

    Four analysts argue in parallel over the top candidates; a judge reads their memos and the data, then returns a verdict and a confidence.

  3. 03

    Construct

    Position limits force an 8–12 name book, then a convex optimiser sets weights under sector and concentration caps.

Every stage is scored separately against realised returns, and the screen is yours to rebuild: drag the factor weights and watch the information coefficient move across 19,051 stock-dates. Open the screen lab →