StockPicker

Describe an investment idea in plain language, turn it into an executable screen, build a portfolio from the results, and put it through an adversarial multi-agent review.

1 · Screen Builder

A language model translates your idea into concrete, numeric criteria. The screen itself runs deterministically in Python, so results are reproducible and every threshold is visible.

2 · Portfolio Builder

Turn screen results into a weighted portfolio — equal, market-cap, score-weighted, or manual — and compare it against a benchmark on return, volatility, drawdown, and tracking error.

3 · Multi-agent review

Three agents argue it out over your holdings, grounded in fundamentals and retrieved SEC filing text:

Pro Agent
Builds the strongest honest case for the position.
Adversarial Agent
Attacks that case claim by claim, not with generic caveats.
Decision Agent
Resolves the disagreement into an actionable call.
Evidence
Claims cite 10-K/10-Q excerpts retrieved via pgvector.
StockPicker provides research tooling, not investment advice. Analysis is generated in part by a language model and can contain errors — verify figures against primary sources before making any investment decision.