
MALI Platform
AI Stock Research
Multi-Agent Debate
Give MALI a ticker and a horizon. Six agents read the same evidence, argue the bull and the bear case, question each other's weakest claims, and a judge works out where the weight of the argument lands.
Plan: EdgeIn the app: AI Intelligence → Agent Debate
What you can do
- Put a stock to six agents, Bull, Bear, Macro, Quant, Sentiment and Risk, over one week, one month or three months, with an optional question of your own.
- Read each agent's thesis, key points and the evidence it cited, and see which agents had no answer.
- Follow the cross-examination: who questioned whom, the weakest claim, whether the agent conceded, and its conviction afterwards.
- Get a verdict with direction, confidence, the probability of that direction, agreement between agents and risk, plus decisive factors, catalysts, risks and dissent.
- See every source the agents were given, the coverage, and why any source was unavailable.
What it doesn't do
- The verdict's numbers are computed from the agents' votes; the language model writes only the summary and can't change them.
- Two agents, Quant and Sentiment, are rules and a sentiment model, not language models, and say so.
- It is research output, not investment advice, and places no orders.
Data: Yahoo daily prices for the stock and SPY, Headlines scored by FinBERT, VIX, rates and the dollar, SEC Form 4 insider trades and 8-K filings, Analyst recommendation counts
Questions
- How long does a debate take?
- One to two minutes. It makes about a dozen model calls and counts as one of the day's AI queries.
- Which plan includes it?
- Edge. Members can run up to 25 debates a day.
- Does MALI remember my context?
- It lists what it remembers about you beside each debate, and you can see and delete it.