Market analyst
Reads price action, trend structure, relative volume and volatility for the symbol you name, and says plainly whether the tape agrees with the story.
Multi-agent equity research
What equity would you like to argue about today?
TradingAgents sends a team of AI analysts at any ticker. They read the filings, the tape, the news and the crowd, then debate each other until one defensible call is left.
Free to try in the console below. Plans add saved reports, watchlists and longer debates.
9 specialist agents12 data feeds per run< 60s from ticker to report4 debate rounds
research console · AAPLIdle
Agent team
Market analyst—
News analyst—
Sentiment analyst—
Fundamentals analyst—
Bull researcher—
Bear researcher—
Research manager—
Trader—
Risk manager—
Type a ticker and run the desk. Nine agents read the filings, the tape and the news, argue both sides, then hand you one report.
Simulated locally — no data leaves this page.
The desk
Nine agents, one call
Each agent owns one question and answers it with its own evidence. Nothing is shared until the debate, which is what stops the team from agreeing on the first plausible story it finds.
News analyst
Scans headlines, filings and transcripts for catalysts inside the holding window, and drops the items that do not move the thesis.
Sentiment analyst
Measures retail chatter and sell-side positioning, so you know whether the crowd is already long before you are.
Fundamentals analyst
Works the last filing: margins, guidance, cash conversion and estimate revisions, with the numbers cited in the report.
Bull researcher
Argues the long case with evidence, then has to defend it against every objection the bear raises.
Bear researcher
Argues the short case just as hard, so a weak thesis falls apart in the debate instead of in your account.
Research manager
Weighs both sides, records which argument won and why, and sets the confidence attached to the rating.
Trader and risk manager
Turns the call into entry, target, stop and position size, then caps the size when volatility or event risk says so.
Workflow
From ticker to a call you can defend
The whole run is written down as it happens, so you can read the reasoning instead of trusting a black box.
Name the symbol
Type any liquid ticker. The desk pulls the tape, the filings, the transcript and the news for that name only.
Agents go to work
Four analysts report first, in parallel, each with its own evidence trail and no shared conclusion to anchor on.
The debate
Bull and bear researchers argue the case, and the research manager records what survived.
One report
You get a rating, a confidence level, entry, target, stop, size and the risk notes behind it.
Coverage
What the desk can read
Depth depends on what actually exists for the symbol. The report always names the feeds it used and the ones that were thin.
| Universe | Symbols | Feeds | Depth |
|---|---|---|---|
| US large cap | 1,000+ | Full filings, transcripts and options data | Deepest |
| US mid and small cap | 3,400+ | Filings and price action, thinner sentiment | Standard |
| ETFs and indices | 600+ | Flows, breadth and sector rotation | Standard |
| Crypto majors | 24/7 | Spot, funding and social velocity | Standard |
Report anatomy
What lands in your account
One page, five parts, no filler. If the desk could not reach a conclusion, the report says that too.
Verdict first
Rating, confidence and horizon sit at the top, so you can stop reading if that is all you needed.
The debate
Both arguments, with the evidence each side leaned on and the point where one of them broke.
The numbers
Valuation, revisions and the levels that matter: entry, target, stop and invalidation.
Risk notes
Concentration, event risk and volatility, including the reasons the desk would step aside.
Sources
Every claim links back to the filing, headline or price series the agent actually read.
Levels, not vibes
Entry, target and stop come out of the same volatility model the risk manager used to size the position.
Every claim sourced
Filings, transcripts and headlines are linked, so you can check what an agent read before you trust what it wrote.
Repeatable
Re-run a name a month later and the new report sits next to the old one, including what changed in the debate.
Why a debate
One model gives you an answer. A desk gives you the argument.
A single prompt finds the first story that fits the data. Structured disagreement is the cheapest way to find out whether that story survives contact with the other side.
Specialists, not one prompt
Each agent has one job, one dataset and no sight of the others until the debate begins.
Confidence, not certainty
Rounds where the two sides stay close are reported as low conviction instead of being smoothed into a buy.
Your runs stay yours
Reports, watchlists and notes are tied to your account and are never published to a shared feed.
Auditable steps
The transcript of the run is kept, so a call can be reviewed after the fact instead of only when it is made.
Built for repeats
Watchlists re-run on a schedule, so the desk tells you when a thesis actually changed.
Explainable by design
Every rating comes with the argument that produced it, which is the part you can learn from.
Plans
Start on the free console, upgrade when it earns it
Researcher, Pro and Team plans add saved reports, watchlists and longer debates. Annual billing takes about 17% off.
Questions
Before you ask the desk
What exactly is TradingAgents?
TradingAgents is an AI research desk. Instead of one model answering one prompt, a team of specialist agents works the same ticker from different angles, argues the case out, and hands you a single report with a rating, levels and risk notes.
Is this investment advice?
No. The output is research material generated by language models from public data. It can be wrong, it can miss context, and it is not a recommendation to buy or sell anything. Size your own risk and do your own checks.
Which symbols are covered?
US large caps get the fullest treatment: filings, transcripts, options and sentiment. Mid and small caps, ETFs, indices and the major crypto pairs run with the feeds that exist for them, and the report says which ones were thin.
How long does a run take?
Most single-name runs finish in under a minute. Longer names with many filings can take a couple of minutes, and the console shows which agent is working the whole time.
Can I keep the reports?
Yes. Every run is saved to your account with its inputs, so you can come back to a call later and check what the desk said before the move happened.
Can I cancel whenever I want?
Yes. Plans are monthly or annual subscriptions billed through Stripe, and you can cancel from your account at any time. Access continues to the end of the period you already paid for.
Give the desk a ticker and see how it argues
Nine agents, one report, and a record of how the call was reached.