All features
Multi-Agent Analysis

Trading Agents

A research desk on every stock — analysts, a debate, a trader, a risk desk, and a portfolio manager.

One prompt gives you one opinion. Trading Agents gives you a process. Click Run Analysis on any symbol and a team of specialised AI agents works it the way a desk would: analysts gather the evidence with live tools, a bull and a bear argue it out, a research manager referees, a trader drafts the plan, three risk profiles stress it, and a portfolio manager signs off — with the full transcript and a written report.

How a Run Works

Seven stages · light model for the volume, deep-reasoning model for the judgment calls

ANALYSTS · TOOLS RESEARCH EXECUTION & RISK VERDICT Market analyst Fundamentals analyst News analyst Sentiment analyst tools: quotes · technicals fundamentals · news · ownership ratings · documents · web search Bull Bear sequential rebuttals N rounds (you choose) Research manager deep tier · stance + plan Trader direction · size · levels risk debate Aggr. Neutral Cons. stress the trader's plan M rounds (you choose) PORTFOLIO MANAGER deep tier · past lessons BUY 72% confidence report · ledger · lessons graded at 1d · 5d · 30d → reflection → fed back into the next run
Light tier — analysts, debates, trader, risk Deep tier — research manager, portfolio manager Typically 12–16 model calls per run; the daily refresher keeps verdicts current between runs without re-running the pipeline.

Full Pipeline or Light Seed

You decide how much thinking each symbol gets

RUN ANALYSIS

The full team, on demand

Every manual run, every escalation the refresher triggers when a thesis flips, and every automatic startup analysis when you choose agents mode: the complete pipeline, with debate depth and risk rounds set in Settings → Trading Agents.

LIGHT SEED

One call to keep every symbol covered

New positions, discoveries, and Refresh All get a single structured call on the light model so nothing sits unanalysed. The daily refresher then revises those seeds only when an input actually moved — and hands a symbol to the full team the moment its call flips from bullish to bearish or back.

Key Capabilities

What the agents do for you

Analysts That Pull Their Own Data

Four analysts — market, fundamentals, news, and social/sentiment — call read-only tools as they work: live quotes, price history and a full technical suite (EMA, RSI, MACD, Bollinger, ATR), fundamentals, your ticker-tagged news, institutional and insider ownership, analyst ratings, and any documents you uploaded. With Deep Research on, the news analyst can also run live Google-grounded web searches.

A Real Debate, Then a Referee

A bull researcher and a bear researcher argue the case in sequential rebuttals for as many rounds as you configure. A research manager on the deep-reasoning tier referees the transcript and must commit to a stance and a plan — no "it depends".

Trader, Then Three Risk Profiles

A trader turns the plan into a concrete proposal — direction, sizing, entry and exit levels. Aggressive, conservative, and neutral risk analysts then stress it in their own debate before a portfolio manager delivers the final BUY / SELL / HOLD / WATCH / TRIM / AVOID with confidence.

Any Provider Drives It

Run the whole pipeline on Claude, Gemini, or ChatGPT. Cheap, high-volume stages run on each provider's light model; the research manager and portfolio manager run on a deep-reasoning model you can choose in Settings. One run is stored per provider, so you can compare how each reads the same stock.

It Learns From Its Own Calls

Every verdict lands in a decision ledger and is graded at 1-, 5-, and 30-day horizons. Resolved calls get a short written reflection, and the portfolio manager is handed the relevant lessons — same ticker and cross-ticker — before it signs off on the next one.

A Report You Can Hand Someone

Every run produces a nine-section written report — analyst briefs, both sides of each debate, the trader's plan, the risk verdict, the final decision — exportable as a branded PDF, and readable from H1 Chat and your MCP-connected assistant.

Related Features

Ready to try it?

Bring your own AI keys and put a research desk on your next decision.

Join Waiting List