Learn

How the analysis happens.

ZenTrading runs a multi-agent pipeline behind every answer. Here's the short version of how it thinks — and a few of the concepts you'll bump into along the way.

How the agent thinks

A four-stage pipeline adapted from the TradingAgents research framework.

Analyze

Fundamentals · Sentiment · News · Technicals

Four specialist analysts gather evidence in parallel — earnings and balance-sheet quality, social and community sentiment, macro headlines and event risk, and technical indicators like RSI and MACD.

Debate

Bullish researcher vs. bearish researcher

Two researchers argue the case both ways. The bull builds the upside thesis from the analysts' reports; the bear stress-tests it. The structured back-and-forth surfaces what each side is leaning on.

Decide

Trader agent

A trader synthesizes the analyst reports and the researchers' debate into a directional view — with timing, conviction, and magnitude. This is where the recommendation actually takes shape.

Risk-check

Risk team · Portfolio manager

A risk team evaluates volatility and liquidity. The portfolio manager signs off on the trade — or sends it back. The output is a clear buy / hold / sell with the reasoning trail intact.

Source framework: TauricResearch/TradingAgents. The hosted agent layers Claude on top of this pipeline.

Most stock chatbots

One model, one pass

A single LLM tries to be the analyst, the bull, the bear, the trader, and the risk officer at once — and usually defaults to a hedged “on one hand, on the other” answer.

ZenTrading

Specialists who actually disagree

Each role plays its own part — the bull and bear are explicitly allowed to argue, the trader has to take a side, and the risk team gets to overrule. You see a recommendation with a real reasoning trail, not a vibe.

Concepts to know

Tap any concept and Claude will explain it in plain English with a real-world example.

Not financial advice. For informational purposes only.