Stop reading twelve tabs.
Ask one AI that has already read them.
EdgeMarket's assistant is wired directly into the warehouse: price history back to 1960, SEC filings, insider and institutional positions, fundamentals, macro and the news feed. Name a ticker — it gathers every block, connects them, and shows you the reasoning.
- Every claim quotes the number behind it, with its date
- It tells you what data it didn't have — nobody else does that
- Callable from any screen — a chart, a filing, a currency, a flow
No card. The terminal is public — the account raises the limits.
The warehouse behind the answer
These counts are read from the database when this page is built and refreshed again in your browser. They are not marketing rounding.
Real chart. Real data. Right now.
Switch the instrument, change the window, turn indicators on and off. This box calls the same public endpoint the terminal calls — if the warehouse is stale, you will see it here first.
Indicators on any series
Moving averages, Bollinger bands, RSI, MACD, volume — computed bar by bar on the server, not approximated in your browser.
Drawing and structure
Trendlines, levels, and the market-structure markers — each one with its measured historical hit rate, not a claim.
The AI reads the same screen
Hit the assistant from a chart and it analyses that instrument, on the numbers you are looking at. Never a second query that disagrees.
This is what an answer looks like
Not a mockup and not a best case — a run that was saved, with the date it was produced on. Check the numbers against that day if you like; that is the point of showing them.
SPY — SPDR S&P 500 ETF Trust
Produced by the EdgeMarket assistant on 2026-08-19, at $769.06. Every number below is quoted from the data pack it was given.
Price is 1.1% below the 52-week high with strong 1-year momentum (+19.5%), but RSI14 at 73.1 signals overbought conditions. The 5.4% drop in institutional holdings is a conflict, and no measured base rates fired, lowering confidence in the directional edge despite the bullish technical trend.
What it weighed
Price 769.06 is 1.1% below the 52-week high 777.88; 1-year return +19.5%; above the 200-day average at 706.73.
RSI14 at 73.1 indicates overbought conditions; 1-week return −0.4% suggests short-term stalling.
13F holdings fell 5.4% to $342.06B in Q1 2026, a conflict with the price uptrend.
10Y yield 4.71% and Fed funds 3.63% create a high-rate ceiling; VIX at 14.89 indicates low volatility.
No measured base rates fired in the last 10 days, providing no historical anchor for direction.
What could break it
- RSI14 at 73.1 suggests a potential short-term pullback due to overbought conditions.
- High 10Y yield of 4.71% may cap further valuation expansion for the S&P 500.
- Lack of measured base rates means the current setup has no statistical precedent to rely on.
Direction and confidence only, with the reasoning shown. The assistant never outputs an entry price, an exit, or a position size, and never sees anything about you — that is a hard constraint of the output schema, not a prompt. This is research, not investment advice.
Three steps, and none of them is "trust us"
It gathers
Eleven blocks of evidence are assembled by calling the same routes the screens call. If a route changes, what the AI sees changes with it — so it can never quote a number you cannot find on screen.
It connects
Insider buying against a falling institutional position. An earnings date against the measured drift after past ones. Rates against a valuation. The connections are the work — and they are printed, not hidden.
It commits — and gets scored
Every directional verdict is stored with the price at that moment and its horizon, then compared with what happened. Confidence bands get a measured hit rate instead of a number the model made up.
Questions people actually ask
What does the AI actually read?
For any ticker it is handed a data pack built from our own warehouse: the full price history and technicals, the latest quote, SEC filing facts parsed from XBRL, Form 4 insider transactions, 13F institutional positions, short interest, company fundamentals, the earnings calendar, macro series and rate levels, and recent headlines. Every number carries its own date, and every block we do not have for that ticker is declared as missing rather than silently left out.
Will it tell me what to buy?
No, and that is deliberate. The output schema has no field for an entry price, an exit, a target or a position size — it is a generation constraint, not a polite instruction the model could ignore. You get a direction, a confidence level, and the full reasoning with the numbers it leaned on. What you do with that is yours.
How do you stop it from making numbers up?
It is only ever given our data, and it is required to quote the figure behind each claim. That rule has a side effect we did not expect: the assistant found a real defect in our own fundamentals — a 720% net margin on NVDA — because a quoted number can be checked and an invented one cannot. We added a plausibility guard and the block now declares itself unusable when it fails.
Is the confidence score real?
It is a three-step label (low / medium / high), never an invented percentage. Language models are badly calibrated on numbers: one that says "80% confident" is right about 55% of the time. Where a measured historical base rate exists for the setup, the assistant is required to anchor to it and quote the sample size and t-statistic — and to say so plainly when no base rate exists.
What does it cost?
A full analysis costs 5 credits, a follow-up question costs 1. Pro includes 1,000 credits a month, which is about 200 analyses. Nothing is charged if the model is unreachable or if the question is off-topic — that check runs before the counter.
Do I need an account to look around?
No. The terminal is open — charts, screener, filings, news and macro are all on screen before you type anything. An account is free and raises the row and history limits; the AI needs credits because it runs on a GPU we pay for by the hour.
The measured signals, when they fire
Which instruments had an event with a measured base rate behind it, and the number that backs it. No market calls, no newsletter filler.
Open it. The data is already on screen.
No card, no demo call, no sales email. The terminal is public — an account raises the limits, and the AI runs on credits.
Free plan for good. Pro is $69.99/mo, or $4.99 for a 7-day trial.See what each plan includes →