Ranked signals across global markets - long or short, refreshed every market close. Teach MANTIS your thinking and it listens, it learns, it refines.
Stop scrolling screens and hunting names by hand. Tell MANTIS what a winning setup looks like once, and it builds the signal and finds every stock that fits - across the whole market, refreshed every session.
Say what a good setup looks like in plain English, or dial in exact parameters - RSI, moving averages, breakouts, volume, gaps, market cap and more.
It turns your edge into a precise signal and runs it across the full stock universe on global exchanges, replaying real price history to compute win-rates and forward returns as it goes.
Get a ranked shortlist of the live stock signals matching your edge - long or short. Every match is paper-traded forward on real prices, so you can prove the edge before risking a cent.
From a quick scan on the phone to a full order ticket on the trading monitor - the same engine and the same paper book, wherever you work.
Each engine stands alone. Together they take a hunch from plain English to a paper-traded, ranked signal.
Describe an edge in plain English or dial in exact parameters - RSI, moving averages, breakouts, volume, gaps, market cap. MANTIS turns it into a precise signal and scans the universe for live matches.
Find a pattern by drawing it freehand, picking one of ten traced reference shapes, snapshotting a window or uploading a chart - then demand the detail that matters, down to a heavy-volume day inside the stretch you marked. Test Lab replays it on real price history with forward-return stats, and on every chart Ask Manti explains the move from real price, news and earnings data, with Bollinger, Keltner and MACD.
An institutional L/S book - net liquidity, long/short exposure, crystallised and monthly P&L, sector and country allocation, advanced metrics. Enter longs or shorts by form or by AI command, paper-traded on real prices.
No synthetic data, no fabricated numbers. Real exchanges, real closes, stated honestly.
MANTIS is not a foundation model - it uses an LLM to turn plain-English edges into precise, testable signals. Every backtest run adds to a growing, dated record of what has held up and what has not, across changing conditions.
Each strategy a user runs adds to a growing, dated log of what was tested and how it actually resolved across changing market conditions - accrued from real usage, not bought in.
The price feed is public - anyone can buy it. What is not public is the accumulating history of which strategies were tried and how they played out. That is created by usage, so it cannot be regenerated from prices alone.
Today it learns your taste and ranks for you; the cross-desk loop - where everyone's testing sharpens the shared filter - is the direction we are building toward. That accumulated, dated record, not the model, is the intended moat.
Jump straight in - the workspace is shared and ready, no account needed.