User Manual

This is the full reference manual for Accessible Trader. It explains every part of the terminal in depth: what each feature does, when you would reach for it, the keys that drive it, and the speech you should expect to hear as you go. It does not assume you already trade: a term is explained the first time it matters, the indicator primer and the risk chapter are written for someone new to markets, and the glossary at the end defines everything in a sentence.

If all you want is a one-line-per-shortcut crib sheet, press F1 inside the app or read SHORTCUTS.md. If you are brand new, the shorter QUICKSTART.md (Quick Start Guide) is a gentler tour. This manual is the place to come when you want to actually operate a feature and understand what it is doing.

How to read this manual

• Keys appear inline, like Alt+K or Alt+Shift+T. A chord such as Alt+Shift+T means hold Alt and Shift together, then press T.
• Spoken feedback is written in quotation marks exactly as the application says it, for example "Loading history…". When a value changes with your symbol or price, it is shown in braces, like "{Symbol} on {Provider}, {Timeframe}. Ready."
• Modifier note: the drawing tools and the detailed point summary are Alt+Shift+letter on every head, desktop and browser alike (since 2026-09-05; the desktop apps used Ctrl+Shift before that). See the Platform Support chapter.

Getting Oriented

Accessible Trader is a full trading terminal — real-time and historical data across stocks, crypto, forex and more — built so that none of it depends on sight. Every function, from placing a drawing tool to tuning an indicator's timbre to submitting an order, is reachable from the keyboard, and nothing requires a mouse. What makes the application different from a conventional platform with a screen reader bolted on is that it was designed around sound from the start, and it treats your screen reader and its own audio engine as two halves of one instrument.

The Hybrid Voice model

The central idea, which everything else builds on, is what the application calls the Hybrid Voice model. Your screen reader and the built-in sonification engine carry different kinds of information at the same time, and neither replaces the other.

Your screen reader — NVDA, JAWS, Narrator, VoiceOver, TalkBack, or Orca on the web host — handles everything exact and textual: the precise open, high, low and close when you land on a candle; an indicator's exact reading when you move to it; dialog labels, menu items and settings; and the confirmation messages after you act. If you need a number, the screen reader is where it comes from.

On the Windows desktop client, the terminal speaks to NVDA and JAWS directly, rather than through the page. That is not a preference, it is a necessity: the chart there is drawn on a native canvas laid over the web view, so while the chart has focus your screen reader is not reading the page at all and anything announced through the page reaches nobody. Whichever of the two is running is used, and one started after the terminal is picked up within a couple of seconds. Nothing needs configuring; if neither is running the terminal falls back to the page, and if that cannot work either it says so in the Journal (Ctrl+Alt+Shift+J) rather than simply going quiet. The Journal also records which of the three routes is carrying speech, which is the first thing to look at if the terminal is ever silent.

On the web host there is no such split — the whole surface is the page — so any screen reader that reads a web page works without special handling.

The sonification engine handles everything continuous and structural: the rising and falling pitch of price as it moves through time, the tonal texture of an oscillator as it swings overbought or oversold, distinct bell tones for events like crossovers and divergences, the stereo position that tells you where in the visible window you are, and volume that scales with how significant a move is. If you need a shape — a trend, a rhythm, a sense of where pressure is building — the audio engine is where it comes from.

The point of running both together is that you perceive precise values and broad structure at once, which is something a conventional chart cannot offer in either direction. During playback you hear them in concert: the screen reader speaks each bar's numbers while the engine plays that bar's sonic shape. You can also lean on one alone — silence the audio and work by speech, or mute speech and listen to pure shape — and later chapters show you how to toggle each as the task demands.

The soundscape

Because so much of the terminal speaks in sound, it is worth learning the vocabulary once; after that it becomes second nature and you will stop having to think about it.

Pitch maps to value, everywhere and consistently. Higher pitch means a higher price or a higher indicator reading; lower pitch means lower. A rising trend is a rising pitch; a sell-off slides downward. For oscillators that have a centre line — an RSI, a MACD — pitch rises above a mid-point for positive readings and falls below it for negative ones, so zero has a recognisable pitch you come to know by feel.

The kind of sound — its timbre, the tonal colour rather than the pitch — tells you where a value sits relative to the structure that matters. Every voice is built on a warm sine base with a slight coloring blended in, and the coloring carries the meaning: an oscillator takes on a brighter square tinge above its midline and a softer triangle warmth below it, so you can hear which side you are on without checking a number; a candle's body adds a deep sub-octave weight that grows with the body's size — a big conviction candle sounds heavier than a doji at the same loudness, so size reads as character rather than volume — and wicks ring as clean pings whose grit grows with their length, the upper wick a bright tone and the lower a deep one, each roughening in proportion to its own reach so a candle with a long tail below and none above is unmistakable by ear. The volume bars sit a perfect fourth below the body — E4 on an up bar, E3 on a down bar, against the body's A4 and A3 — so the bed is audibly under the candle rather than on it, and neither lands on a wick. Until 2.11.0 volume played on the body's own pitch, which is why a busy bar could sound like one thick note instead of two. And when a reading pushes into an overbought or oversold extreme — on a bounded oscillator such as RSI, Stochastics, MFI, CCI, Williams %R, or the Ultimate Oscillator — a noise texture roughens the tone so the extremity itself is audible. It is a pronounced roughness, not a faint wash, and you can set how strong it is per level with the Zone Texture slider in an indicator's properties (P), under Reference Levels. You learn to hear "deep oversold" the way a sighted trader sees a line pinned to the bottom of its range.

On top of the continuous tones, discrete events ring as short bells, and each event type has its own bell so you can tell them apart by ear: a smooth sine bell for crossovers, a bright metallic triangle bell for divergences, a high pure crystal bell for support and resistance levels, a shimmering detuned pair for high-confluence signals, and a rich multi-harmonic blend for momentum signals. When a bell catches your attention during playback you can stop and read the exact event through your screen reader.

Reference levels themselves (the 30/70 lines on an RSI, a zero line, your own levels) speak up in three stages as a value interacts with them. Approaching a level — within about five percent, not yet across — fires one very short, very high ping (1,400 hertz), louder the closer the value sits to the line; it is a "heads up, almost there" cue and fires once per approach. Crossing the level rings a two-note chirp: rising pitch for an upward cross, falling for a downward one. And when a value has stayed beyond the level for several bars, a single low, longer tone (220 hertz) confirms "still out there". These are navigation cues, not trade signals — they are not written to the Journal, and none of them changes with your strategy setup. If you hear one lone high beep while arrowing across an RSI, that is the approach ping.

Adding a level of your own. Press 0 with a series focused and the terminal drops a reference line on it. What it does depends on which pane you are on, because a level only means something in the units of the thing it sits on.

On the price chart there is no meaningful constant at all, so the line goes at the price of the bar under your cursor: arrow to the level you care about, press 0, and you have marked it.

On an oscillator the line goes on that pane's neutral — the value the reading actually swings about. For MACD, momentum and the other zero-centred readings that is zero. For RSI, Stochastic and MFI, which run 0 to 100, it is 50. For Williams %R, which runs −100 to 0, it is −50. This used to be zero everywhere, which put a line on the floor of every 0–100 oscillator — a value RSI never visits, so a line that could never be crossed and could never make a sound. Where the indicator already declares its own midline — RSI ships one at 50 — you are told so and nothing is added, because two lines at one value would report every crossing twice. And on a pane where nothing declares a neutral, the key says so rather than guessing.

You will hear which one you got — "Zero line added", "Midpoint added at 50" or "Level added at 63,920.11". From then on that line speaks and pings like any other, so you can hear price approach and cross the level you chose — from either direction, straight away, without visiting any settings. Ctrl+Left and Ctrl+Right jump to those crossings too: the midline is one of the places that key stops, alongside the overbought and oversold lines.

Removing one. Press 0 again on the same bar and the level comes off; you will hear "Level removed". Levels an indicator declared for itself — an RSI's 30 and 70 lines — are never removed this way, because they are part of what the indicator is; those live in Properties. Properties also gives every level a Remove button, a colour, a line style, a thickness, and a choice of which crossings to report: either direction, only rising through, or only falling through. All of it is saved with the workspace and survives a restart, and "Reset to defaults" restores the indicator's own levels while leaving yours alone.

Two more cues round out the picture. Stereo position places you in time: the leftmost visible bar plays hard in the left channel and the rightmost hard in the right, so as playback advances or you move the cursor rightward the sound travels across the stereo field and you feel your position in the window without a spoken word. And the engine layers sound by depth — quieter background tones for long-term context, mid-level tones for the main price and oscillator voices, and the loudest layer reserved for signals and event bells — so an important event cuts through the continuous wash instead of competing with it at the same volume.

None of this needs to be memorised before you start. Load a chart, press the right arrow a few times, and let the pitch move with the price; the rest of the vocabulary arrives naturally as you meet it in the chapters that follow.

Loading a Market

When the terminal first opens it is deliberately quiet. There is no chart, no symbol, and no spoken greeting — just an empty workspace waiting for you to tell it what to look at. That silence is intentional: the application never assumes which market you trade or whose data you pay for. This chapter takes you from that blank start to a live, navigable chart, and explains what you will hear at each step so you always know where you are in the process.

The whole sequence is short once it is familiar: connect a data source if the provider needs credentials, pick a market, provider, symbol and timeframe from the toolbar, and press Load. The terminal then fetches recent history, announces that the chart is ready, and — where the provider supports it — begins streaming live bars.

Connecting a data source

Market data comes from providers, and many of them require an account and an API key before they will return anything. You manage those credentials in the API key manager, which you open with Alt+K. Your screen reader announces the dialog as "API Keys", and it opens on a list of any profiles you have already configured, followed by a form for adding a new one.

A profile is one set of credentials for one provider in one environment. The "Add New Profile" form walks top to bottom: choose the Provider (Alpaca, Binance, Coinbase, Gemini, Kraken, Kraken Futures, Oanda, Polygon, Schwab and the rest, or "Custom" — note that Kraken Futures is its own venue, not a setting on Kraken: it has a different host, different request signing, and API keys you mint separately, so it needs its own profile), give the profile a Profile Name you will recognise later — the placeholder suggests something like "Alpaca Paper" — and set the Environment to either Paper or Live. Paper points the provider at its simulated/sandbox endpoints; Live uses your real, funded account. Below that you set the Market Type (Spot, Futures, Crypto or Stocks) and then enter the API Key, the API Secret, and, only if your provider issues one, a Passphrase (it is labelled Optional and you can leave it blank for providers that do not use one). All three secret fields are masked.

Everything the profile stores — the secrets and the profile list itself (which providers you use, the nicknames, the environments) — is held in your operating system's encrypted credential storage, never in a plain-text file on disk. Older versions kept the profile list (though never the secrets) in a plain file; the first run of a current build migrates it into encrypted storage automatically and removes the old file.

Activate the Save Profile button and the terminal confirms with "Profile {name} saved". A profile must be the active one for its provider before its data or trading access is used in the current session; if you keep more than one profile for a provider — say a Paper profile and a Live profile — select the one you want and activate it, and you will hear "{name} set as active". The active profile is read back in bold in the profile list so you can confirm at a glance which environment you are about to trade against. This Paper/Live distinction matters later when you place real orders, so it is worth getting into the habit of checking it here first.

Schwab is the one provider that does not take a typed key, because it uses a browser sign-in instead. For a Schwab profile you will see a Sign in button; activating it announces "Opening Schwab sign-in in browser", hands you off to Schwab's own login in your default browser, and on return confirms "Schwab sign-in complete for {name}".

Not every provider needs a key. Several crypto sources and the free historical archives chart happily with no credentials at all, so if you only want to study those you can skip this dialog entirely and go straight to the toolbar. The terminal will tell you, in the next step, on the rare occasion a key is actually missing — you do not have to memorise which providers need one.

The toolbar, in two rows

The toolbar has two rows and they divide by purpose, which is worth learning once because it tells you where to look for anything.

The first row is the buttons. Left to right it holds the object tree and the sound designer; then the trading dashboard, order book, strategies, Watch lists (and the screener), Levels (the respect report), Trade journal and AI analyst; then alerts and API keys; then save and load workspace; then settings and help. After those come the controls that change how the chart is shown — pan left and right, zoom in and out, and the four display toggles: heatmap, Heikin Ashi, log scale and Replay. The last four are pressed-state toggles, so your screen reader announces whether each is currently on. The things you put on a chart — Add indicator, Drawing tools and Scripts — sit on the bar under the chart instead. Every button that opens a dialog has a keyboard shortcut named in its tooltip, so the toolbar is how you find a feature and the shortcut is how you reach it once you know it is there.

While a dialog is open, the rest of the terminal is switched off — not merely covered over. The toolbar, the tab bar, the chart, the indicator bar, the status bar and the footer stop taking focus and disappear from your screen reader's view of the page entirely, so Tab cannot walk out of the dialog and a stray review command cannot land you on a toolbar button behind it. Escape brings all of it back, and focus with it: the chart if that was the last dialog, or the dialog underneath if you had two open. The one thing that keeps speaking is the terminal's own announcements — those live outside the switched-off region on purpose.

The second row builds the chart: the market cascade described below, then import (for your own data) and Load chart.

The buttons are icons. Each one's name is what your screen reader announces and what its tooltip shows on hover, so nothing is lost by the icons carrying no text; what was gained is the height — with a caption under every icon the toolbar took three lines on an ordinary window and the chart was left less than half of it. If you would rather see the names on screen, Settings → Appearance → Toolbar captions puts a caption under every icon again. Some buttons only appear when they apply — drawing tools and the chart toggles are hidden on analytics charts, where they mean nothing.

Focus mode

Press Alt+Z, or the Focus mode button on the bar under the chart, and the toolbar, the tab bar, the indicator bar and the footer disappear; the chart takes their height. The status line stays, because it is the visual mirror of what was just spoken, and it gains an Exit focus mode button so there is always a visible way back. The terminal announces both directions and names the chord when it turns on, since the controls it just hid were also tab stops. Every shortcut still works in focus mode; only the bands are gone. It is not saved — the next start opens with everything showing.

Choosing what to chart

Everything you select to build a chart lives on the toolbar's second row, in the order the terminal needs it: Market, Provider, an optional Type, Symbol, and Time. There is no dedicated shortcut for these fields; you Tab into the toolbar and Tab through them left to right, and your screen reader reads each control's label and current value as you land on it.

These four selectors form a cascade — each choice decides what the next one can offer. Choosing the Market (for example Crypto, Stock, or Forex) refills the Provider list with the sources that cover that market and automatically selects the first of them. Choosing a Provider refills the Symbol list and the available timeframes and, again, selects the first symbol for you.

The Market list also holds one special entry, Analytics, which is where you chart data feeds rather than tradeable instruments. There is no longer a separate Trading/Analytics switch — picking Analytics in the Market dropdown is how you cross over. When you do, an extra Analytics type selector appears right after Market offering Economic, OnChain, Derivatives, and Sentiment; choose one and the Provider and Symbol lists refill with that category's sources (FRED economic series, on-chain metrics, funding/open-interest, Fear & Greed, weekly CFTC fund positioning, daily short-sale volume for any US stock, and so on). For everything else — actual markets you can trade — you simply never touch the Analytics entry. One consequence worth understanding: moving through this cascade does not speak on its own — the terminal repopulates the dropdowns silently, and it is your screen reader, reading each list as you open it with the arrow keys, that tells you what is now available. So after picking a market, expect to arrow through the Provider and Symbol lists to hear and confirm your choices rather than waiting for an announcement.

Some providers split a market into a Type — most commonly Spot versus Futures on crypto exchanges. When that distinction applies, a Type selector appears between Provider and Symbol; when it does not, the field is simply absent and the Tab order closes up around it.

The Symbol list is a standard dropdown rather than a search box. Open it and type the first letters of a ticker to jump to it — typing "B" then "T" walks you toward BTC pairs, for instance — exactly as you would in any list your screen reader knows. This is also where a missing credential surfaces: if the provider you picked requires a key you have not configured, the Symbol list contains a single entry reading "⚠ API key required — open API Keys (Alt+K)", and the Load button stays disabled until you go back and add one. That sentinel is the terminal's way of pointing you to Alt+K at exactly the moment it matters.

Time is two controls working together: a multiplier you type (1 to 999) and a unit you choose — min, hr, day, wk, or mo. Together they read as a timeframe such as 1 hr or 15 min; the default is 1 hr. When a provider advertises a set of common timeframes, quick-pick buttons appear at the end of the row, after Load chart, so you can jump straight to, for example, "Set timeframe to 1h" without touching the multiplier — and so that Tab reaches Load right after the unit, not after a dozen pills. A quick-pick sets the timeframe; Load still loads it. The quick-picks are always the provider's own list — a button you can press is a timeframe that provider genuinely serves. Two consequences follow. When a provider offers exactly one timeframe (most analytics feeds — Fear and Greed, COT positioning, short volume — are daily-only), the whole Time area disappears: there is nothing to choose, and the timeframe still reads in the tab title and the Shift+F1 context summary. And when you switch to a provider that doesn't offer your current timeframe, the terminal snaps to one it does offer and tells you so — "provides 1d data only; timeframe set to 1d" — instead of silently fetching nothing.

A worked path makes the cascade concrete. Suppose you want Bitcoin against the dollar on Binance at the hourly. Tab to Market and select Crypto; the Provider list fills and lands on Binance. Tab to Provider and confirm Binance (or arrow to another exchange). Tab past the Type field — leaving it on Spot — to Symbol, open the list, type "BTC" to reach BTC/USDT, and select it. The Time field is already 1 hr, so you are done choosing. Because the cascade pre-selected sensible firsts, a common shortcut is to pick only the market, glance through to confirm the provider and symbol it chose for you, and load that.

Loading the chart

With a symbol chosen, Tab to the Load button and activate it. The terminal announces "Loading history…", interrupting whatever was being said, while it fetches the most recent bars. When the data is in and the chart is built, it announces the full identity of what you are now looking at — "{Symbol} on {Provider}, {Timeframe}. Ready.", for example "BTC/USDT on Binance, 1h. Ready." — again interrupting so you hear it immediately. That "Ready." is your cue that the chart is populated and every navigation and playback command in the rest of this manual is now live.

If the fetch fails — a bad or inactive key, a provider outage, an unsupported symbol — you hear "Chart failed to load." instead, and the toolbar shows a matching error such as "Chart load failed. Check provider settings." The usual fixes are to confirm the right profile is active (Alt+K) and that the symbol is one the provider actually serves.

At any later moment you can re-hear what is loaded without touching the toolbar: press Shift+F1 and the terminal announces the current symbol, provider, and timeframe. This is handy after you have been deep in navigation for a while and want to reconfirm the instrument before acting on it.

Once you are used to the toolbar you rarely need to Tab back to it. Ctrl+Alt+Shift+L loads the chart from wherever you are — it does exactly what activating the Load button does. If you have not yet chosen enough to load anything it says so rather than doing nothing quietly: "Cannot load yet. Choose a market, provider, and symbol first." That makes it safe to reach for out of habit; the worst case is a sentence telling you what is still missing.

Knowing whether your data is live

A loaded chart is not automatically a live one. After the "Ready." announcement the terminal keeps working in the background: it fills in any gap between the history it fetched and the present moment, and then, if the provider streams, it switches to receiving real-time ticks. There is no separate "you are live now" banner, so the terminal tells you in the language it uses everywhere else — sound and speech.

The clearest signal is a closing bar. Once a live stream is running, each time the current bar closes and a new one opens you hear a short bell and an announcement in the shape "Close {price}. {pattern}. New bar: Open {price}." — for example "Close 42,500. Doji. New bar: Open 42,510." Those rolling new-bar announcements are the sound of a live feed; their arrival is your confirmation that data is flowing. (If you find them distracting you can switch them off in settings, so treat their presence, not their absence, as the positive signal.)

To check the live edge deliberately, press Backslash (\) to jump the cursor to the most recent bar. If you navigate forward past the end of the fetched data into time that has not happened yet, the terminal says "No live data yet", which tells you you have run off the end of what exists rather than that anything is wrong. A provider that offers only historical data — or a live connection that has not come up — simply never produces those new-bar bells; the chart stays perfectly usable for study, it just is not advancing on its own.

Switching markets later

You are not locked into your first choice. Return to the toolbar at any time, change any selector, and press Load again to replace the chart. One case is worth a heads-up: if you load an analytics provider — chosen through the Analytics entry in the Market dropdown, one that returns single scalar metrics like an economic series rather than OHLCV candles — onto a tab that already holds indicators or drawings, those tools cannot apply to a non-candle series, so the terminal stops to confirm with a "Switching to analytics" dialog. It offers three choices: "Continue (strip & load)" replaces the chart and removes the tools that no longer fit, "Open in New Tab" loads the analytics series beside your existing work and leaves it untouched, and "Cancel" backs out. When in doubt, open it in a new tab so you keep both views.

With a chart loaded and confirmed live, you are ready to read the market itself — which is where the next chapter, on moving through time bar by bar, begins.

Your own data: the My Data market

The Market dropdown has one entry that isn't a market at all: My Data — your own CSV files, charted with everything the terminal knows how to do. Select it and the Symbol list shows what you've imported; an Import data file button appears beside the cascade (or press Ctrl+Alt+Shift+I from anywhere). The import dialog takes a file or — often easier with a screen reader — pasted CSV text, and it tells you exactly what it understood before anything loads: "Imported Household Budget: value series, 240 rows, January 2005 to June 2026."

Three file shapes are recognized from the header row. A file with date,open,high,low,close (volume optional) charts as full candles — playback, pattern announcements, even backtesting your own history. A file with a date column plus any named number columns — date,Income,Expenses,Net — makes each column its own loadable chart ("Budget — Income", "Budget — Expenses"). And a file shaped date,label,value becomes event markers: add "My Events: your-dataset" from the indicator dialog (Alt+A, category My Data) and each event lands on the bar covering its date, spoken with its own label — put your real fills on the chart where they happened and hear "Bought 0.5 BTC, 42,000" as you arrow past. Dates read best as 2026-07-22; common spreadsheet formats and unix timestamps also work, and the dialog has copyable templates for all three shapes. Excel and LibreOffice users: File, Save As, CSV. If a file can't be read safely the import is refused with the line number and reason — a refused import beats a silently wrong chart.

Imported data doesn't have to live on its own chart. The indicator dialog's My Data category puts any dataset ON the chart you're viewing, aligned to its bars (sparser data holds its value between points, like the weekly COT series does on a daily chart). "My Data: X" opens the dataset in its own pane with every column navigable, and a Normalize-to-100 setting makes different-sized columns comparable by shape. "My Data overlay: X" draws it on the price pane itself, rebased so it starts where the chart was — from there the two lines diverge by relative performance, in pitch as well as pixels. For imported OHLCV data, "My Data ratio: X" adds the classic strength read: the chart's close divided by yours, rising when the loaded symbol outperforms.

The same comparison works against any exchange symbol without importing anything: the indicator dialog's Overlays category has "Compare symbol (overlay)" and "Compare symbol (ratio)". Type the symbol (and optionally a different provider — it defaults to the chart's own), and the second market appears rebased on the price pane or as a strength ratio in its own pane, always on the chart's timeframe so the bars line up one to one.

How a Chart Is Laid Out

Every key in the next chapter moves through a structure. Sighted traders absorb that structure in one glance and never name it; this chapter names it, because by ear you meet it one piece at a time. If you already know what a price chart looks like, skim the headings and read the part about components, levels and zones, which is specific to this terminal.

A picture you can build at home

Take a sheet of paper and lay it on the desk in landscape, the long edge toward you. That sheet is one pane. Run a ruler along its bottom edge, left to right: that is the X axis, and it is time. The far left is the oldest bar you have loaded, the far right is now, and every centimetre is one bar of the timeframe you chose: one minute, one hour, one day. Stand a second ruler upright along the right-hand edge: that is the Y axis, and on the top sheet it is price, low at the bottom, high at the top.

Now stand a row of matchsticks on the sheet, one per bar, each one placed where its price sits on the upright ruler. Each matchstick is a candle. Its thick body runs from the opening price to the closing price; the thin ends sticking out above and below are the wicks, the highest and lowest prices touched during that bar. A long body is a decisive bar; a body so short the candle is almost all wick is a bar where buyers and sellers ended about where they started. Lay a length of thread across the tops of the matchsticks and you have drawn the shape of price through time. That thread is what pitch plays for you: high on the ruler is a high note, low is a low note, and the sound sweeps from your left speaker to your right as the cursor moves along the bottom ruler.

Put a second sheet of paper directly below the first. It has the same bottom ruler, because time is shared, but its own upright ruler with its own units. This sheet might be volume, counted in shares or coins rather than dollars. A third sheet below that might be RSI, a gauge that only ever reads between 0 and 100. Each sheet is a pane, and the whole stack, sharing one timeline, is the chart.

A pane is a Y axis

That is the whole definition. Things share a pane when they are measured in the same units, and get their own pane when they are not.

• The Main pane is price. Candles live there, and so does every overlay measured in price: moving averages, Bollinger Bands, VWAP, Ichimoku, pivot levels. They sit on the same sheet because a moving average at 42,500 belongs beside a candle at 42,500.
• Volume has its own pane, because a volume of 1.2 million is not a price.
• Every oscillator has its own pane. RSI reads 0 to 100. MACD on Bitcoin swings by hundreds of dollars. Put them on one upright ruler and RSI's entire working range becomes a sliver at the bottom of MACD's, and by ear that sliver is one flat note. Until September 2026 thirty indicators shared one pane and that is exactly what happened; now each gets a pane of its own, and two copies of the same indicator share one because they are on the same scale.

Alt+PageUp and Alt+PageDown move between panes and name the one you land on. They stop at the ends with the boundary tone rather than wrapping, because a silent jump from the bottom of the chart back to the top is the one move you could not detect.

Bounded panes keep their scale. An RSI pane always runs 0 to 100, at every zoom. An unbounded pane, MACD or ATR, fits itself to the values in view. Seventeen indicators declare their natural range and four declare a floor, and the difference matters by ear: RSI 70 is the same note on every chart, whereas a MACD reading is a note relative to whatever else is in the window. Alt+L switches the price pane's ruler to logarithmic, so equal distance means equal percentage, which is the honest ruler for a history in which price multiplied. The ear follows the same ruler: a level-pitched component in the price pane, the close line for one, takes its pitch from where the value sits on the scale as drawn, so on a log chart the price halfway up the pane is the note halfway up the sweep. Indicator panes are never on the log scale, on screen or in sound.

"Halfway up the sweep" means halfway in PITCH, not halfway in hertz. From 2.12.0 a value's height in its pane maps to the 200–1000 Hz band exponentially, so equal fractions of the pane are equal musical intervals wherever in the pane you are: a tenth of the pane is always the same step, about two and a third semitones, whether you are down near the low of the window or up at the high. It used to be a straight line in hertz, which sounds like a straight line only if you hear in hertz — the bottom quarter of every pane carried a whole octave and the top quarter carried a major third, so a price riding high in the window moved a long way on screen and barely at all in the ear. The floor and the ceiling of the band are unchanged, so the extremes sound exactly as they always did; it is the middle that has been redistributed, and the midpoint of a pane now sounds at 447 Hz where it used to sound at 600.

What the price pane fits, and why it is your choice (Alt+F). By default the price pane sizes itself to hold everything drawn on it: the candles and every visible overlay. A Bollinger band or a Keltner channel that swings wider than the bars widens the pane rather than running off it. That is not only a picture question, because the pane's range is also the pitch range — anything outside the pane is not merely invisible, it is silent, and a band that has gone quiet sounds exactly like a band sitting still. Press Alt+F and the pane fits the price bars alone; the terminal says "fitting price only". Overlays may then leave the pane, and go quiet when they do, but the price line has the whole 200–1000 Hz band to itself, which is what you want when you are listening for small moves in price rather than watching a band. Press Alt+F again to go back. It is per tab, it is saved with the workspace, and a new tab inherits it.

The trade is worth stating plainly, because it is the reason this is a switch and not a decision made for you. Adding a wide band to the price pane compresses the price line into a smaller share of the pane, and therefore into a smaller share of the pitch band. Under the interval-even mapping above that compression is a constant transposition — a price line occupying a third of the pane gets a third of the octaves, wherever in the pane it sits — so it is a cost you can predict and reason about rather than one that changes depending on where price happens to be standing.

Series, components, strips

Inside a pane the structure continues. A series is one thing you added: the candles, an RSI, a trend line. A component is one line or marker that series draws, and it has a value on every bar. MACD is one series with three components: the MACD line, the signal line and the histogram. Cipher B is one series with eleven. PageUp and PageDown walk series; Up and Down walk the components of the series you are on.

A few panes divide into strips, a band of the sheet with its own upright ruler: Cipher B keeps its money-flow histogram in a strip along the bottom of its pane. Ctrl+Up and Ctrl+Down walk the strip you are in across every series drawn in it, which on the Main pane is how you step from the candles to an overlay drawn on top of them.

Components, reference levels and zones

Three different kinds of thing live on a pane. They are heard differently and navigated differently, and telling them apart is most of what this chapter is for.

A component has a value per bar. An EMA has a value on every bar; a buy dot has a value only on the bars it fired. You navigate onto a component and hear its reading; in playback it plays as a tone (a line) or a bell (a marker). Components are what Up, Down, Ctrl+Up and Ctrl+Down move between.

A reference level is a constant. RSI's 70 and 30, a Stochastic's 80 and 20, MACD's zero, ADX's 25: one number, the same on every bar, drawn as a dashed line across the pane. On the paper picture it is a thread stretched taut across the sheet at one height. A level has three properties the terminal keeps track of:

• its value, which you can move in Properties;
• its meaning: overbought, oversold, or neutral (the line the reading swings about), or a named band on either side: ADX at 25 separates "developing trend" from "strong trend", and Choppiness at 61.8 separates "trending" from "ranging", which is inverted, a low reading meaning a trend;
• its owner: the indicator declared it, or you added it with the 0 key. Yours can be removed with 0 again; the indicator's live in Properties and come back with "Reset to defaults".

You do not navigate onto a level. You hear it when a component interacts with it: a very short high ping as the value comes within about five percent, a two-note chirp as it crosses (rising for up, falling for down), and a single low tone once it has stayed beyond the line for several bars. Ctrl+Left and Ctrl+Right stop at the crossings, and on a band-labelled line the announcement names the band you entered. A level you hide in Properties stops being all of those things: no ping, no chirp, no jump target.

A zone is a region, not a line. Three kinds appear:

• the shaded overbought and oversold bands behind an oscillator, which are visual only and never spoken or navigated, though the noise texture you hear inside them is the audible version of the same idea;
• zone lines, which Cipher SR and Spider Lines carry forward from past turning points as support and resistance. These are levels with a side, and they get the full vocabulary: "support at 61,200 broken", "price tested resistance at 64,000, tested 3 times", "approaching support at 61,200";
• value zones from the Value Deviation indicator: bars where price turned back toward value, tiered by how far out it was, marked with a glyph and a pitch per tier.

Drawings are none of these: a trend line you placed is its own series, read value-first when you arrow along it, and announced when price crosses it.

Hearing the whole thing

Three keys describe the structure instead of a value.

• Alt+Shift+/ describes the pane you are in: "Main pane, 1 of 3. Y axis, price: 41,200 to 43,800, about 500 between gridlines, linear scale. X axis, time: March 3 2026 to July 1 2026, 120 bars at 1d each." It then names the series in the pane, and on a pane with strips says "Control Up and Control Down walk the strip you are in."
• Ctrl+Alt+Shift+Y describes the chart: the symbol and timeframe, how many bars are in view of how many loaded, the price range and scale, how many indicator panes sit below the price pane, each with its own Y axis, how many components are hidden or muted, whether Heikin-Ashi is on, and whether the feed is fresh.
• Shift+F1 says where you are: symbol, provider, timeframe, the focused series, its pane and where that pane sits in the stack.

The numbers these three speak come from the same ranges the chart is drawn with, so what you hear is what a sighted partner sees.

Heikin-Ashi, and what stays raw

A standard candle reports what happened in one bar: it opened here, traded up to there and down to there, closed here. A Heikin-Ashi candle (Alt+C) is a smoothed version: its close is the average of the bar's open, high, low and close, and its open is the midpoint of the previous smoothed candle. The effect is that a trend becomes a run of same-coloured candles with wicks on one side only, and choppy noise inside the trend is averaged away, which is why the sequence of pitches is easier to follow with HA on.

The price of that smoothness is that a Heikin-Ashi close is not a price anything traded at. So the terminal keeps two things raw whatever the candles are doing:

• Every indicator computes on the raw bars. RSI, MACD, a moving average, a volume profile: all read the real open, high, low and close. Switching HA on does not change a single indicator reading, and a level the indicator announces is a level in real prices.
• Chart formations, the close line and the title price are raw. A neckline measured from a smoothed candle could not be put into an order, so formations are read from standard candles even while HA is displayed, and the terminal says so when you switch it on with formation description enabled.

What does change with HA on is the candle you hear: the spoken open, high, low and close, the body and wick percentages, and the candle pattern named on the bar are all the smoothed candle's, because they are meant to describe the shape on screen. Use HA to hear the trend; trust the levels as real.

Market Watch and Screening

Loading one symbol at a time answers "what is this market doing?". Market watch answers the question before it: "which market should I be looking at?" It holds your watchlists, and it runs a screen — the same conditions a strategy uses for entry, asked across every symbol on a list at one instant instead of across every bar of one symbol.

Open it with Alt+M, or with the Watch button on the toolbar's first row. The dialog has three tabs — Watchlists, Build a screen, and Run screener — and you move between them with Tab and Space like any other tab list.

Watchlists

A watchlist is a named, ordered set of symbols, each remembering which provider and sub-type it came from. Order is yours: nothing re-sorts behind your back, and Up and Down on each row move a symbol explicitly.

Create one by typing a name into New list name and pressing Create list. The Add a symbol area below is a cascade, exactly like the toolbar's: Market, then Provider, then Sub-type, then Symbol. Choosing a market narrows the providers; choosing a provider loads its real symbol list, so you pick from what actually exists rather than typing a symbol name and hoping. When the dialog opens, all four are pre-filled from the chart you were looking at, which makes "add this to a list" a single press of Add.

Providers can list thousands of symbols, so there is a Filter symbols box beside the picker. Type any fragment — usdt, EUR, SOL — and the list narrows as you type; a live message under the picker reads out how many symbols are showing and how many exist in total. That message also tells you when the list is capped at 500 and you need to narrow further, so a truncated list never poses as a complete one. Add all shown adds everything the filter is currently showing, and announces both how many were added and how many were already on the list.

Every row on the list has Load, which puts that symbol on the chart through exactly the same path the toolbar's Load button uses, then closes the dialog.

Building a screen

A screen is a set of filters plus a rule for combining them. On the Build a screen tab, press Add filter and you get one row with four or five controls:

• Indicator — which indicator the condition reads. Split from the component deliberately: one flat list of every signal in the app runs to several hundred entries and is miserable to move through by ear.
• Component — which of that indicator's outputs, for example Cipher B's Buy signal or RSI's Value line.
• Condition — what has to be true. The choices offered depend on what kind of component you picked, and that gating matters: a marker component has no value on bars where it did not fire, so "is above 30" would be false forever and would read to you as a quiet market rather than as a nonsense filter. A marker offers fired on the last bar and fired within N bars; an oscillator offers thresholds, ranges, crosses, and percentiles; a cloud offers inside/above/below; a level offers rejection and break.
• Value, Upper, Within bars, Weight — only the boxes the chosen condition actually uses appear, and a fresh filter is seeded with a threshold somewhere sensible inside the component's own range rather than at zero.

Under each row, a plain-language sentence restates the whole filter — "Cipher B — Buy fired within 3 bars." — so you can check a row in one read instead of tabbing back through five controls.

Combine filters decides the logic. All must be true is a strict AND. Any may be true is an OR. Weighted score reaches a threshold gives every filter a Weight box and matches when the weights of the filters that are true add up to the threshold you set — useful when you want confluence ("any three of these five") rather than unanimity.

Bars of history per symbol is the one setting that quietly breaks screens if you get it wrong. Every indicator needs a warm-up before it produces values, and if you fetch fewer bars than the slowest indicator you referenced needs, every symbol comes back as "not enough history". The default of 500 covers everything that ships; raise it if you build a filter on a long moving average.

Save screen stores it under the name you gave it, and selects it in the Run screener tab so you can run it immediately. Editing an existing screen is the same dialog — pick it from Edit saved screen, change what you like, save again. If a screen was built elsewhere and contains nested groups, the flat builder says so rather than showing you a plausible-looking subset, because saving a flattened copy over it would destroy structure you cannot get back.

Running a screen

The Run screener tab takes three things: which saved screen (or none, which matches everything and gives you a plain quote board), which watchlist to run it against, and on what timeframe. Press Run screen.

Progress is spoken as it goes — how many symbols are done, and which one is in flight — and Cancel stops it. Results arrive as a real table with proper column and row headers, so your screen reader reads the column name with each value as you move cell by cell. Each row carries the symbol, provider, whether it matched, the score if the screen used weights, last close, percentage change, a status, and a Load button that puts it on the chart.

Two things about the results are deliberate. First, symbols that could not be evaluated are reported, never dropped: a row reading "not enough history" or a fetch error stays visible even with Show matches only ticked, because "we could not check twelve of these" must not be able to look like "nothing qualified". Second, the spoken summary after every run names all three numbers — matched, evaluated, and failed — for the same reason.

Reading the Chart

A loaded chart is a structure you move through, not a picture you glance at. This chapter covers how you navigate that structure by keyboard, how you scan it for the moments that matter, how you play it back as sound, and how you control what you hear while you do. None of it requires sight; all of it is faster once the handful of movement keys are in your fingers.

The shape of a chart

The chart is a stack of panes sharing one timeline; a pane is a Y axis; a pane holds series, a series holds components, and a level or a zone is a different kind of thing from either. The previous chapter, How a Chart Is Laid Out, builds that picture from scratch. Hold it in mind here: you change panes one way, series another, components a third, and the terminal names where you have landed each time.

Moving through time

Left and Right arrow move the cursor one bar at a time — Left into the past, Right toward the present. As you land on each bar your screen reader announces its values: on the price pane, the open, high, low and close of that candle; on an indicator component, that component's reading for the bar. At the same time the sonification engine plays the bar's tone, so a run of Right-arrows is also a little rising or falling melody of where price went.

When you want to cover ground faster, Home jumps to the leftmost bar in view and End to the rightmost, while Backslash (\) leaps all the way to the latest, live bar. You can also move the window itself without moving the cursor: the bracket keys pan the viewport, with [ bringing older bars into view and ] newer ones. Shift+[ makes each of those presses move a smaller distance and Shift+] a larger one, which matters more than it sounds: on a daily chart you usually want to step a week at a time, and on a one-minute chart a week is thousands of bars away. Set the step to suit the timeframe once and every later [ or ] lands where you expected. The minus key zooms out to see more bars at once, and the equals key zooms in for finer detail on fewer.

If you prefer the mouse, the chart toolbar carries Pan left, Pan right, Zoom in, and Zoom out buttons that do exactly the same things and speak the new visible range just as the keys do, so they are safe to use with a screen reader. You can also click and drag the chart itself to pan — provided no drawing tool is selected — dragging right to reveal older bars; letting go of the button anywhere stops the pan.

Using the mouse on the chart

Every mouse action on the chart lands in the same place the keyboard navigates, so the mouse and the keyboard always agree about where you are — and everything a mouse action does is spoken through the same pipeline the arrow keys use. That makes the mouse fully usable alongside a screen reader, and it makes the chart approachable for a sighted or low-vision trader sharing the terminal with you.

Click a bar to hear it. A single click on the chart moves the reading cursor to the bar under the pointer and announces it exactly as if you had arrowed there — values spoken, tone played. Because the click really moves the cursor, you can click roughly in the area you care about and then fine-tune with Left and Right arrows; pointing precision is never required. Clicks in the empty space to the right of the newest bar do nothing.

Click near an indicator to focus it. If your click lands close to an indicator line — an EMA, a MACD line, a band — keyboard focus also moves to that series and component before the bar is announced, so what you hear is the thing you pointed at. A click that isn't near anything specific simply selects the bar on whatever was already focused, so imprecise pointing still works.

Shift+click to measure. Hold Shift and click a bar, and the terminal speaks a range summary from the reading cursor to the clicked bar — how many bars, the dates, the high and low of the span, and the net change in price and percent. Measuring never moves your cursor, so you never lose your place.

Magnet snap for drawings. Off by default; toggle it from the chart's right-click menu. With magnet on, drawing anchors pull to the nearest open, high, low, or close of the bar under the pointer when you get close — trend lines land exactly on the wick or the close without pixel-perfect aim.

Hover sound. Also off by default, also in the right-click menu: a soft, short tick as the pointer crosses each bar, pitched to that bar's closing price — sweep the mouse across the chart and it hums the price contour, without touching your reading cursor or interrupting speech.

Scroll wheel. Scrolling zooms in and out, keeping the bar under the pointer fixed in place. Shift+scroll pans through time instead — down or right for newer bars, up or left for older — using the same pan step as the bracket keys, with no button-holding or dragging needed, which makes it the easiest way to move through history if clicking and dragging is difficult for you. A sideways swipe on a trackpad pans the same way.

Double-click to jump to now. Double-clicking anywhere on the chart jumps straight to the newest, live bar and announces it — the mouse twin of the Backslash key.

The crosshair. As the pointer moves over the chart, a crosshair follows it: a vertical line snapped to the bar under the pointer, a horizontal line at the pointer's price, and a readout in the top corner showing the bar's date, the price at the pointer, and the bar's open/high/low/close. The readout is real text on the page — not part of the chart image — so screen magnifiers, browser zoom, and custom styles all work on it. It never speaks (that would be constant chatter as the mouse moves); the spoken equivalent is simply clicking the bar. You can turn the crosshair off and on from the chart's right-click menu.

The right-click menu. Right-click on open chart space (or press the Application key — or Shift+F10 — with the chart focused) for the chart menu: Play from here starts playback at the bar you clicked; Jump to latest returns to the live edge; Show/Hide crosshair, Magnet snap, and Hover sound toggles; and beneath those, every series on the chart is listed by name. Better still: right-click near an indicator line and the menu opens directly on that series' actions — no hunting the list — with Back one press away. Choose a series to get its actions — Focus, Mute or Unmute, Hide or Show, Properties, and Remove. Listing series as menu items is deliberate: acting on an indicator never requires clicking on a thin line, which matters as much for shaky hands and low vision as it does for screen reader users. Right-clicking directly on a drawing's anchor handle still opens the drawing's own menu (Delete, Duplicate, Properties), as before.

Touch and mobile

On a touchscreen — a phone or tablet browsing accessibletrader.com, or a touch-screen laptop — the chart understands the gestures you would expect, and each one lands in the same place the keyboard navigates, announced through the same speech pipeline:

• Tap a bar to move the reading cursor there and hear it, exactly like clicking.
• Drag with one finger to pan through time, like grabbing the chart.
• Pinch to zoom in and out, anchored between your fingers.
• Double-tap to jump to the newest, live bar.
• Press and hold for about half a second to open the chart menu (or a drawing's menu if your finger is on its anchor handle).

On touch devices a navigation toolbar also appears below the chart with large, plainly-labelled buttons — Previous bar, Next bar, Previous component, Next component, Previous series, Next series, Place drawing point, Play, and Chart menu. If a gesture ever misbehaves, the buttons always work, and they are the most reliable way to drive the chart with VoiceOver or TalkBack running (swipe to the button, double-tap to press). Previous/Next series are the touch equivalents of Page Up/Down, so you can move between loaded series without a keyboard. Place drawing point is what makes drawing tools usable by touch alone: arm a tool from the chart menu's Drawing Tools, move the cursor with the bar buttons or a tap, then press Place drawing point to drop each anchor — no keyboard shortcut needed. It speaks a hint if no tool is armed.

With a mobile screen reader, the screen reader owns the touchscreen, so the chart also offers a bar navigator — announced as "Bar navigator" just before the chart. It behaves like a slider: focus it and flick up or down (VoiceOver) or swipe up/down or use the volume keys (TalkBack) to step through bars, each one spoken with its position, date, and closing price, with the full announcement and tone following from the app's own speech. One honest caveat: iOS VoiceOver moves web sliders in steps of about 10% of the chart rather than one bar at a time — flick to get close, then use the toolbar's Previous/Next bar buttons for single steps. TalkBack on Android steps one bar at a time. Finer VoiceOver control arrives with a future native iOS update.

Touch in the installed mobile apps flows through the same layer and is expected to behave the same way, but it has not yet been verified on physical devices — until it is, a connected Bluetooth keyboard remains the fully-supported way to drive the mobile apps. Pinch-zooming the page itself is no longer blocked anywhere, so browser-level magnification works on the website.

Explore mode turns the touchscreen into a reading surface. The touch toolbar's Explore button (announced with its state) switches a single finger's drag from panning to exploring: slide across the chart and each bar under your finger speaks — value first, then the date — with a pitch tick that traces the price contour, and the crosshair follows for sighted partners. Lift to stop; touch again anywhere to keep reading; pinch still zooms mid-explore. Screen reader users: activate the Explore button through the toolbar as usual, then use your screen reader's pass-through gesture on the chart surface to hand it your finger. Turning Explore off restores drag-to-pan, and the terminal says so.

Moving between panes and components

Four keys, four levels of the structure. Page Down and Page Up move to the next or previous series in the order the chart draws them, top to bottom; as you arrive speech names it — "RSI", "Volume". Up and Down step through the focused series' components: Down from the MACD line to the signal line to the histogram, each announced with its name and current value. Alt+PageDown and Alt+PageUp move between panes, the next Y axis down or up the chart, and the pane's name is spoken at the end of the move so you hear what you landed on after you hear what it says. Ctrl+Down and Ctrl+Up walk the strip you are in across every series drawn against the same axis — from the candles to a price overlay drawn on top of them, which is the pair you most often want to compare.

All four clamp at the ends with the boundary tone; none of them wraps, because a silent jump from the bottom of the chart back to the top is the one move you could not detect.

Three keys re-orient you whenever you lose the thread. Shift+F1 announces the symbol, provider and timeframe, the focused series, and its pane and place in the stack. Ctrl+Alt+Shift+C focuses the chart and reads a fuller context summary. Ctrl+Alt+Shift+Y describes the chart's layout rather than its values, and Alt+Shift+/ describes the pane you are in: what each axis measures, its range, the gridline step, and what is drawn there. Reach for all of them freely; there is no penalty for asking the terminal where you are.

Ctrl+PageUp and Ctrl+PageDown are left unbound on every head. Browsers use them to cycle their own tabs ahead of anything the page can do, so pane navigation lives on Alt+PageUp and Alt+PageDown everywhere and the desktop and the browser agree. F1 always shows the bindings actually in effect on the host you are using.

Scanning for events

Stepping bar by bar is precise but slow when what you actually want is the next thing that happened. Ctrl+Left and Ctrl+Right are context-aware jumps that depend on what you have focused. On a price candle they jump to the previous or next bar where price crosses a trendline you have drawn. On an oscillator that crosses zero, like MACD, they jump to the next zero-crossing; on a banded oscillator like RSI, to the next entry into or exit from overbought or oversold; on a moving-average overlay, to the next price-versus-average cross; on a sparse signal marker, to the next bar where that signal fires. Press repeatedly to walk through every such event in turn; when there are no more in that direction the terminal says "No more {component} signals in this direction." It turns a long history into a short list of the moments worth hearing.

When a pane carries more than one reference line, the jump uses the line the focused component answers to, not the first line in the pane. Aroon is the case worth knowing: Up and Down swing about a Midpoint at 50 while the Oscillator swings about zero, so with the Oscillator focused the key lands on its zero cross and says so. Before 2.11.0 it landed where AroonUp crossed 50 and announced that as the Oscillator's Midpoint cross — both halves of the sentence naming the wrong thing. Past a component's only line you now hear "No crossing in view" rather than a sibling's crossing offered as your own.

Playback: listening to the market

Playback is the core listening mode: the cursor animates through the visible window while the engine plays each bar's sound and your screen reader speaks its values, so you hear the shape of a stretch of market in seconds. Three keys set how much you hear. Space plays or stops the whole chart — every visible, unmuted series at once, each sounding all of its own visible, unmuted components together, so a busy chart really does play as a full ensemble, down to the soft cloud- and ribbon-fill washes (an EMA Fill between two averages, say) that used to drop out when too many voices were in play. Shift with Space plays just the series you have focused, all of its components, for studying one indicator without the rest. Ctrl+Shift+Space narrows further to a single component — the RSI line alone, say. Muting a series or component with M, or hiding it with H, drops it straight out of the mix, which is how you thin a crowded soundscape down to what matters.

Those two single-key toggles are fast on purpose, and that makes them easy to over-use: hide four components across three indicators and there is no practical way to walk back and find them, because a hidden component is exactly the thing you can no longer navigate to. So there are two undo-alls. Ctrl+Alt+Shift+K brings back everything you have hidden and Ctrl+Alt+Shift+U unmutes everything you have muted, each announcing how many it restored — "3 items shown.", or "Nothing was hidden." if there was nothing to bring back. Narration has the same escape hatch: Ctrl+Alt+Shift+O switches narration off on every series and clears every component you picked out with N, so a selection built up across three indicators over an afternoon is one chord from a known state — "Narration off for 2 series.", or "Nothing was narrating." Treat all three as the way out whenever the chart has gone quieter than you meant it to, and Ctrl+Alt+Shift+Y (above) as the way to hear what is currently hidden or muted before you decide. Ctrl+Space pauses and resumes whatever is playing — and while paused it now falls properly silent instead of holding the last chord, though the arrow keys still audition individual bars so you can inspect the frozen moment. Shift+Escape is the panic key that stops all playback at once, and Shift+= speeds playback up while Shift+- slows it down — slower to dwell on each bar, faster to scan a long history.

What playback says while it runs. The tones carry the price, so the words carry everything the tones cannot. Three things speak, all of them without interrupting each other and all of them composed into a single sentence per bar so nothing gets cut off half way:

• Where you are in time. Each time the bars cross a calendar boundary one step coarser than the bar spacing — a new hour on minute bars, a new month on daily ones — the new period is spoken: "February 2024." The unit is chosen so the announcements land at least two seconds apart at whatever speed you are playing, so speeding up makes them coarser rather than more frequent.
• Signals, on the series you asked for. A marker signal printing on the bar the tones have just reached is spoken — "Triple confluence buy, strong confirmation." This is the same flag as live auto-narration, and there are three questions behind what you hear: N picks what MAY speak, the Narration tab picks WHEN, and the scope you played picks WHICH of them. Press Space and the whole chart may speak; Shift+Space on one series and only that series does. Only discrete events, never a running commentary — at ten bars a second a per-bar readout would be a wall of speech.
• A line crossing one of its own levels, which is discrete in exactly the same way: an ADX crossing 25 happens a handful of times in five hundred bars, not on every bar. A line that declares what its two sides mean says the band you entered — "strong trend" — and anything else names the line it crossed. These go through the same rarity ranking and the same ceiling as markers, so a line that whipsaws across its level is treated as the routine thing it is.
• When several fire at once, the rarest is the one you hear. Two clauses is the ceiling for one bar, and what falls off the end is the marker that fires most often on the chart in front of you. That matters on an indicator like Cipher B, whose gold Triple Confluence dot can only print on a bar that is also an oversold crossover and also a WaveTrend cross: all three fire together, and the one you added the indicator for is the one that prints twice in four hundred bars. The rare one leads the sentence too, because at ten bars a second the first clause is the one that lands before the next bar sounds.
• A formation resolving. If you have "Describe chart patterns" on, a chart formation whose story ends on the bar being played is spoken there, in the same words the arrow keys would use if you stopped on it.

Signals and formations are rate-limited to the same two-second cadence as the landmarks: a second one arriving inside the window is dropped, not queued, because at ten bars a second a queue means hearing about a bar the tones passed eight seconds ago. One thing gets through that window anyway: a rarer signal than the one that opened it. A WaveTrend cross every dozen bars would otherwise swallow the gold dot four bars behind it, which is the opposite of what the limit is for. It cannot cascade — whatever speaks claims the window at its own rarity, so the next thing has to be rarer still. Landmarks are never dropped by that limit — they are the only thing telling you where in time you are. The master switch is Settings (F12) → Narration → Narrate during playback: turn it off and playback is tones and nothing else, with the start, pause, speed and finish confirmations still spoken so the end of a run never sounds like a crash. The landmarks have a switch of their own beneath it — Speak time landmarks during playback, on by default — because the date and the signals answer different questions, and wanting to hear what your indicators printed is not the same as wanting the calendar read to you every few seconds.

Volume profiles do not narrate during playback. VPVR, VPFR and TPO announce their point of control, value area and the point of control moving at a bar close with auto-narration on, and they say nothing while playback runs. That is deliberate rather than missing: a profile's levels are recomputed as the profile itself rebuilds, so unlike a fixed line at 25 there is no stable answer to "how often does price cross this" — and that count is what the rarity ranking uses to keep playback from becoming a per-bar readout. Play the chart for the tones and read the profile with the arrow keys, or leave auto-narration on and let the bar closes tell you.

A signal is introduced by the component that fired it, never by the series. "Bullish divergence." — and where the wording does not already say which marker it was, the component leads: "WaveTrend Cross Bull: Wave cross up 12." You chose which series narrate, so the indicator's name tells you nothing you did not know; which of its markers just printed is the fact you are waiting for.

What playback does NOT say, and where to find it instead. A continuous line — a moving average, a VWAP, an oscillator — never speaks during playback, whether or not you have flagged it with N. It has a value on every bar, so narrating it would mean a number ten times a second for the length of the run, and the tones already carry it. The events those overlays produce are narrated on the bar close instead, where there is a whole bar interval for the words to land in: price crossing your EMA is announced there, as is an oscillator changing zone. So the division is not a special case for playback so much as one rule seen from two sides — playback speaks what happened at a point, bar-close narration speaks what changed.

If you want playback as pure narration rather than as sound, F3 turns the chart tones off and leaves everything above running: the cursor still walks the chart, the words still arrive, and nothing plays underneath them.

Choosing what you hear

You are always in command of the output layers, and the F-key row follows one rule: the plain key controls what you asked for, and Shift controls what happens to you. F2 toggles interactive speech — navigation values, zoom and pan announcements, summaries; with it off, your commands run silently. Shift+F2 toggles event speech — alerts, monitoring reports, new-bar announcements. F3 toggles chart sonification — the navigation tones and playback; Shift+F3 toggles earcons. Two things refuse to be silenced: errors, and your order outcomes — fills, stop hits, take profits speak and sound through every mute, because missing a stop firing costs real money. (If you truly want total silence, Settings → Speech has an "Event mutes also silence order fills and stops" switch — read its warning first.) Individual alerts can also be marked Break through mutes when you create them, for the handful that must never be missed. F4 toggles the braille display on platforms that support one, and Shift+F4 opens Settings on its Braille tab, with focus already on it. None of these mutes persist — the terminal always starts with everything audible.

One extra control exists on the public website versions (the demo and the hosted terminal), where the last speech hop is your browser: Speech output on this device. A browser cannot tell whether a screen reader is running, so on your first visit the terminal asks — let your screen reader do the talking, have the browser's voice read everything aloud, or both. Answering "screen reader" is what stops the doubled voice some Chrome users hear (Orca and the browser voice saying everything twice). The choice is remembered per browser, and can be changed any time in Settings under Speech. Volume is adjustable at three depths: F5 and Shift+F5 raise and lower the focused component, F6 and Shift+F6 the whole focused series, and F7 and Shift+F7 the master chart volume — so you can bring one quiet line forward without touching the rest. The M key mutes or unmutes the focused series or component without removing it, and H hides or shows it; both are toggles, and both leave the data in place so you can bring it back with the same key.

Text labels on the chart

Alt+Shift+L pins a text label at the cursor bar. The anchor goes down first, then a small dialog asks what the label should say — so the position and the wording are two separate decisions, which matters when you cannot see where the anchor landed.

The text is drawn on the chart beside the anchor, and it is read wherever you meet the label. Arrow or jump onto the bar it is pinned to — from the label's own series or from any other — and you hear a short high two-note tick followed by the wording. The tick is deliberately unlike any tone the chart makes from data: a label is an annotation, not a measurement, and it plays no price tone at all. The wording is also the label's name, so the object tree lists it and the legend shows it. A label reading "Label (3)" would be a label you have to go and look at, which is the one thing that does not work here.

Off its bar, the label's own series says the wording and "not on this bar" — a label is a single point, so most bars of its series are empty and silence there would be indistinguishable from a series that had stopped working.

Cancelling the dialog leaves the label in place with no text, and says so — deleting something you had just deliberately positioned would be the worse surprise. Remove it from the object tree if you did not want it.

Bar replay

Playback reads you a chart you can already see all of. Bar replay does the opposite: it hides everything after a point in history and gives it back one bar at a time, so you meet the market the way you meet it live — without knowing what happens next. It is the honest way to practise, and it is the only way to find out whether a setup you like the look of was actually readable at the time or only looks obvious with the rest of the chart in view.

Start it with Ctrl+Alt+Shift+P (or F11 on the desktop), or with the Replay button at the end of the toolbar's button row. Replay begins at the bar your cursor is on, and everything after it disappears.

On the web host use Ctrl+Alt+Shift+P rather than F11 — browsers own F11 for fullscreen and will not pass it through.

While replay is running, the chart is a normal chart in every other respect. Indicators recompute on the revealed bars only, so an oscillator reads exactly what it would have read at that moment; you can navigate, inspect a bar, draw, and place paper orders. Stopping replay restores the full history and the viewport you had before you started.

The Replay toolbar button shows its own state, so you can tell at a glance — or by the button's announced pressed state — whether history is currently hidden.

Inspecting a single bar

When one bar deserves a thorough look before you act on it, press Alt+Shift+D for a full point analysis. It reads the candle's open, high, low, close and volume, names any candlestick pattern recognised at that bar — "Engulfing bullish" — reports every active indicator's reading there, and lists any signal events on the bar across all indicators. It is the one-key way to gather everything the terminal knows about a single moment, which is exactly what you want in front of a decision — and a natural lead-in to the analysis tools in the next chapter.

Analysis Tools

The chart is only the canvas. This chapter covers what you put on it to make sense of the market: indicators that compute and sonify their own readings, drawing tools you place by keyboard and then hear the cursor cross, the volume profile that maps where trading actually concentrated, and the heatmap that colours activity into the playback itself.

An indicator primer

An indicator is arithmetic over the bars, drawn as one or more lines or markers. None of them knows anything the candles do not; each one compresses the candles into a shape that is easier to read for one particular question. Four families cover most of what you will add.

Trend: where is price relative to its recent average? A moving average is the mean close over the last N bars, redrawn each bar, so it lags price and smooths it. A simple moving average (SMA) weights every bar equally; an exponential moving average (EMA) weights recent bars more, so it turns sooner. Price above a rising average is the textbook uptrend; a fast average crossing a slow one is the textbook signal. VWAP is the average price weighted by volume, the level the day's trading actually centred on. ADX measures trend strength without direction: below 20 no trend, above 25 a strong one, above 50 a very strong one, which is why its pane names those bands.

Momentum: how fast, and is it stretching? These are the oscillators, and they get their own panes. RSI (relative strength index) compares recent gains to recent losses on a 0 to 100 scale: above 70 is called overbought, below 30 oversold, and 50 is neutral. Overbought does not mean "sell"; it means the last fourteen bars were mostly up, and strong trends stay overbought for weeks. Stochastic asks where the close sits inside the recent range, 0 to 100, with 80 and 20 as its lines. MACD is the gap between a fast and a slow EMA, drawn as a line that swings about zero, a smoothed signal line, and a histogram of the difference between them; a cross of zero is the two averages crossing each other. Rate of change is simply the percent move over N bars.

Volatility: how much does it move? ATR (average true range) is the average size of a bar in price units, and it is the number stops and targets should be measured in, because a 1% stop is tight on one instrument and enormous on another. Bollinger Bands draw a moving average with a band two standard deviations either side; the bands pinch when the market is quiet and open when it is not. Keltner and Donchian channels are the same idea built on ATR and on recent highs and lows.

Volume: who is participating? The volume bars themselves. OBV and the accumulation distribution line add volume up on up bars and down on down bars, so a rising line with a flat price says buying is building. MFI is RSI weighted by volume. A volume profile turns the question sideways: instead of volume per bar, volume per price level, so you can hear which prices the market accepted (the point of control, where the most traded) and which it passed through quickly.

Multi-signal and structure. Cipher B combines a wave oscillator, money flow and divergence detection into one pane with eleven components and rare gold "triple confluence" dots. Market Structure labels swing highs and lows as higher or lower than the last and reports the trend state those labels imply. Value Deviation marks where price turned away from value. All three are descriptive: they tell you where you are, not what to do.

Two cautions travel with every indicator. First, warm-up: an indicator with a fourteen-bar period has nothing to say for its first fourteen bars, and a screen built on a long moving average needs enough history to produce a value at all. Second, every indicator reads the raw candles, whatever Heikin-Ashi is doing to the picture; the section on Heikin-Ashi in the chart-layout chapter explains what that means.

The glossary at the end of this manual defines each of these terms in a sentence.

Adding and tuning indicators

Press Alt+A to open the Add Indicator dialog. Indicators are grouped into categories — Multi-Signal, Trend, Momentum, Cycles, Positioning, Derivatives, Volatility, Volume, Profiles, Overlays (comparison symbols) and My Data — and you move through the category and indicator lists with the arrow keys and add one with Enter. The dialog says whether the indicator joins the price pane or gets a pane of its own. A new indicator arrives with audio properties already chosen for its type, so it is immediately playable; you can refine them later.

Drawing tools are not in this dialog, and never should have been. Trend lines, Fibonacci levels, the measure tool, the risk/reward tool and the rest are placed with Alt+D (Drawing Tools) or with their own shortcut chords, because a drawing is a thing you position on the chart rather than a calculation you switch on. Until September 2026 fifteen of them were also listed here, and adding one from this dialog produced a series with nothing in it — no points to place and nothing to hear. They have been removed from the list; Alt+D is the way.

Some indicator lines have new names, and five indicators have gone. Until August 2026 a number of indicators drew nothing at all: they were listed, they could be added, and they produced an empty line with no error — indistinguishable, by ear, from a market with nothing to say. Two separate causes, both now fixed and both now guarded by tests.

Working again, unchanged in how you use them: Bollinger Bands, Keltner Channel, Chandelier Exit, Ultimate Oscillator and Momentum. If you added any of these before and assumed you had mis-set something, you had not.

Renamed lines, because the old names were the reason those lines were blank. The indicator behaves the same; only what speech calls the component has changed:

TRIX, ROC and ADL also gained a smoothing period in their settings. Their signal and average lines were declared but could never be filled in, because the setting that controls them was not offered; the lines work now and default to sensible periods.

No longer listed: PPO, ZLEMA, TMA, Historical Volatility and Ease of Movement. The maths library the terminal uses does not implement these, so they could never have drawn anything. They have been withdrawn rather than left in the list — an indicator you can add and wait for is worse than one that is honestly absent. They return if the library gains them.

Once it is on the chart you reach it the same way you reach any series — Page Down from the price pane, or Alt+PageDown to its pane — and explore its components with Up and Down, as the previous chapter described. Two small touches help here. Pressing 0 on a focused oscillator toggles the pane's declared neutral line (50 on RSI, zero on MACD), which sounds as the reading crosses it and gives Ctrl+Left and Ctrl+Right a place to stop; the soundscape section in Getting Oriented has the full rule. And when an indicator has outlived its use, Delete removes the focused series after a spoken confirmation, and Ctrl+Z brings it back — the candles themselves cannot be removed.

To change how an indicator calculates or sounds, focus it and press P (or Shift+F12) for its properties dialog. It has five tabs: General for the calculation parameters — periods, smoothing, thresholds; Appearance for colours and line styles; Levels for every reference level, its value, colour, which crossings to report, and a Remove button on the ones you added; Sonification for the sound patch each component plays, with a bullish/bearish or above/below-midline pair where the component has two sides, and the zone texture; and Speech for how the series reads. A "Save as Defaults" option stores your preferences so the next indicator of that type starts already configured the way you like, and "Reset to Defaults" restores the indicator's own settings while leaving your levels alone.

Positioning indicators — who actually holds what. Two 1.6.0 additions read official positioning data instead of price. COT Positioning (Positioning category) speaks the weekly CFTC report — hedge-fund net position as a 26-week z-score, with "crowded long" and "crowded short" bells at the ±1.5-sigma extremes — and picks the right futures contract from whatever chart you are on (gold, silver, copper, oil, gas, Bitcoin, Ether, the S&P, the Nasdaq, the euro, the dollar index). Its detail facts tell you honestly where the signal has tested well (contrarian on gold, a dip-buy gate on equity indices) and where it has not (CME crypto, FX). For individual stocks, load the FINRA analytics provider and chart {TICKER}_SHORTVOL — the share of each day's volume that was sold short, a daily crowding gauge for any US equity. Both sources are free and need no key.

A note on Cipher A. As of 1.6.0 Cipher A is retired from the Add Indicator dialog: its engine is the same WaveTrend as Cipher B, so it added no independent information. Saved workspaces and strategies that use it keep working exactly as before — it is hidden from the menu, not removed from the terminal. Tab and the arrow keys move through the dialog and your screen reader reads every label and value.

Two whole-chart toggles live near the indicators. Alt+C switches the price pane to Heikin-Ashi candles, and Alt+L to a logarithmic price scale; both are explained in the chart-layout chapter, including the one thing to hold on to about Heikin-Ashi — every indicator keeps reading the raw candles.

Market structure

Market Structure is the answer to "where am I?" before any indicator answers "what should I do?". A new chart starts without it — candles, volume and price only — and you add it from Alt+A like any other indicator, or switch it on for every new chart under Settings, General, Analysis. It finds the swing highs and lows, labels each one against the one before it — higher high, higher low, lower high, lower low — and reports the trend state those labels imply.

A swing high is a bar whose high is the highest of the bars either side of it, within the Pivot span (five bars each side by default), and a swing low is the mirror. The Minimum swing size (ATR) setting suppresses pivots that are too small to be structure rather than noise — a swing has to differ from the previous one by at least one ATR by default.

The indicator marks a red square on each swing high and a green square on each swing low — just clear of the wick, so the price they mark stays readable — and rings a ping for each as you navigate over it, a low tone for lows and a higher one for highs. Two further events get their own marks, both drawn as an X: an amber X for a Break of Structure, when price closes beyond the last swing in the direction the trend was already going, which is continuation; and a larger purple X for a Change of Character, when it closes beyond the last swing against the trend, which is the first mechanical evidence that the trend may be over.

The shapes are deliberate. Market Structure owns the angular family — squares and crosses — and Value Deviation below owns triangles, dots and diamonds, so that on a chart carrying both you can never mistake a swing high for a resistance zone. They shipped a week apart, were each checked alone, and for a while both drew red down-triangles at the same size. Three more components are computed but hidden by default — Structure State, Last Swing High and Last Swing Low — and you can switch any of them on in the object tree or the indicator's properties.

One thing to be clear about, because it is easy to misread. A swing marker is drawn ON the pivot bar, but a pivot cannot be identified until the span has passed — five more bars have to fail to exceed it. So the mark appears in a place you could not have traded: by the time it shows up, price has already moved five bars away from it. A chart full of triangles sitting exactly on the lows is the most seductive illusion in technical analysis. Testing this directly, filling at the pivot price versus filling at the close of the bar where the pivot could first be known, was the difference between an impossible return and one that did not beat buy-and-hold. Use structure to understand where you are, not as an entry trigger.

The indicator is descriptive by design and it says so in its own description. You can turn it off for good in Settings if you would rather add it per chart.

Chart formations

Chart formations are the multi-bar shapes a sighted trader names in one glance — double tops, head and shoulders, triangles, wedges, flags, ranges. Delivering that by ear is the reason this terminal exists, so the terminal describes them; it never tells you what they mean. Turn it on in Settings → General → Describe chart patterns. It is off by default, because it is extra narration on an action you perform constantly. The separate Describe candle patterns switch beside it covers the one-to-three-bar shapes and is on by default — it rides on announcements you already asked for rather than adding an occasion to speak. Since 2026-09-04 it governs the arrow keys as well: turning it off leaves the candle reading as "Bullish" or "Bearish" and the prices, which is what it said before the multi-bar patterns arrived.

Twelve shapes are recognised: double top and double bottom, head and shoulders and inverse head and shoulders, ascending / descending / symmetrical triangle, rising and falling wedge, bull and bear flag, and the range (a flat top against a flat bottom — the most common state a market is in).

The three things that can happen, and how you hear each one

Every formation has a trigger level — one price that decides its fate. On a double top or a head and shoulders that price is the neckline; on a triangle or wedge it is the boundary a break would cross first; on a flag it is the flag edge. A range is the one exception and has two.

The wording is deliberately literal. You will never hear the word "completed", because it could not tell you whether the pattern worked or failed — it only ever meant "price closed through a line". So the terminal says which side of which level, and leaves the meaning to you.

Where you are in the shape

A formation is a stretch of chart, not a single bar, and the announcement tells you which edge you just crossed:

• On its first bar: "Start of possible double top forming… Spans 22 bars." The outcome has not happened yet, and the neckline is the number to remember.
• On the bar where it resolved: "End of double top: price closed below the neckline…", or when reached going forward, "Double top confirmed here…" / "…ends here without confirming."
• Between those points: silence. It has already been described and nothing has changed.

The edge word describes the bar, not the direction you came from. A formation's first bar says "Start of" whether you arrowed onto it going left or right, and its last bar says "End of" either way. That is the only arrangement in which the readout is a reliable map: if the word changed with your direction of travel, the same bar would describe itself differently depending on how you got there, and you could not build a picture of the chart by moving around in it.

Press , and . to jump between formation edges — the start of each shape and the bar its story ended. Two keys walk you through every formation on the chart in the order they happened. They only work while formation description is on; with it off they say so rather than moving you somewhere without explaining why.

Press Alt+Shift+D on any bar for the full list, including every overlapping formation with its own levels. If nothing is live where you are standing, it tells you what finished most recently and how — "No formation here. Most recent, 20 bars ago: double top: price closed below the neckline at 42,100." That broken level is often still the most relevant price on the screen.

Reading a break versus a hold — the part that is easy to get backwards

Take the double top, because it is the clearest case and the confusion is common.

A double top is two highs at roughly the same level with a trough between them. The neckline is that trough, and it is support. So:

• Price closes below the neckline. The double top has confirmed. In the conventional reading this is the bearish case and the pattern working as advertised: the level that was holding price up gave way. The textbook entry is short on the break, or on a retest of the broken neckline from underneath; the measured target is the neckline minus the height from the twin tops down to it. A break is the pattern succeeding, not failing — the shape is a top, and tops are supposed to break downward.
• Price comes to the neckline and holds. The double top failed to confirm. Support was tested and it survived. Conventionally that is the bullish case, and the level that held is the reference an upside trade is built around.

Both are useful and they point opposite ways, which is exactly why the terminal reports what price did rather than whether the pattern "worked". Every shape follows the same logic once you know which side its trigger sits on, and the announcement always names the side:

Playing the second top. There is a second, earlier trade hiding in the forming announcement, and it is worth naming because it is the reason forming patterns are reported at all. Once you hear "possible double top forming" you are being told that price has returned to a level it was rejected from once already. Acting there — at the second top, before the neckline is anywhere near — is the anticipation trade: a much better price, a much tighter stop just above the twin highs, and a much lower chance of being right, because most possible double tops never become double tops. Waiting for the neckline break is the confirmation trade: worse price, wider stop, higher hit rate. The terminal gives you both moments and takes no view on which one to take.

Ranges

A range is announced with both boundaries — "Possible range forming, top 110, bottom 100. Height 10." — and no target, because a range that has not broken has not chosen a direction and projecting one would be inventing an opinion the shape does not hold.

When it goes, you hear which way: "Range breaks here: closed above the top at 110, measured target 120." Until then the two numbers are the whole content: the conventional readings — buy the bottom, sell the top, or wait for the break — all need the same two prices, and the terminal gives them to you without choosing between them.

If it expires still intact you hear "Range ends here still intact — price held between 100 and 110", which is a different statement from a failed pattern and is worded differently on purpose.

Overlap: when several shapes fit at once

A stretch of chart can genuinely be an inverse head and shoulders and a double bottom and an ascending triangle at the same time. Two experienced traders looking at it would disagree about which it is, and the terminal is not going to pretend otherwise by silently picking one.

What it does instead is rank them and describe the leader, then count the rest: "…Plus 2 more formations here." The ranking is live formations before resolved ones, then the largest structure first — the eighty-bar shape is what the chart is making, the twelve-bar flag inside it is a detail of that shape. Size is the tie-break because it is the only one available that is not a directional opinion; ranking by "which pattern is more reliable" would be exactly the untested claim this terminal refuses to make.

Alt+Shift+D reads them all when you want the disagreement in full.

A formation inside a larger one says so. "…Inside a larger double bottom that began 12 March." The container's start date is given rather than just its name, because that is what lets you go and find it — and it is the difference between a setup that stands on its own and one that is a detail of a shape still in play. Where shapes nest three deep the immediate parent is named, not the outermost, because that is the level you are actually standing in.

And you can choose which one leads. Press ; to cycle through the overlapping formations at the current bar; the one you pick is described first from then on, on that chart, until you press Shift+; to go back to largest-first. The default ranking exists because size is the only ordering that is not a directional opinion — but your setup may well be built on the small one, and the terminal has no business insisting otherwise. Nothing is hidden either way.

The cycle covers every formation over the bar you are standing on, container and contained alike — a range, the double bottom inside it, and the flag inside that are three presses apart, and a fourth press wraps back to the first. Each announcement says which of how many you are on ("2 of 3"), so you always know the size of the set you are walking.

While one is chosen, , and . walk that formation's edges only — its start and its ending, nothing else. Choosing a shape and then being carried to a different one's break bar is the behaviour this replaced, and it made the choice feel as though it had not registered: you would hear "leading with ascending triangle" and, one keypress later, "double bottom confirmed here." Both were true. The key had simply gone somewhere you did not ask to go. When you reach the end of a chosen formation the terminal names it and reminds you that Shift+; releases it.

What the terminal will not tell you, and why

You will never hear a formation called bullish or bearish, and you will never hear a probability. The measured target is spoken because it is arithmetic on two numbers already on your screen, and it is always phrased as the measured target, conditional on if it breaks — that is the difference between reporting a convention and endorsing one.

That reticence is not caution for its own sake. Every price-derived pattern claim this project has tested has come back null: a randomly-drawn horizontal line was respected 59% of the time, real swing levels held 46.2% of the time against 46.7% for random lines, and fib ratios did nothing across 355,000 tests. The shapes are real and worth hearing — they are how you build a picture of the chart. Whether acting on them makes money is a separate question, and one this project has repeatedly failed to answer in their favour. Use them to understand where you are, and get your edge somewhere you have tested.

Seeing them, if that helps

Settings → Appearance → Draw chart formations. Off by default. It shades each formation's span, draws the trigger level solid, and draws the conventional measured target faint and dashed.

The audience is anyone who is not listening: a low-vision user, a sighted trading partner, a screenshot in a bug report. If you are working by ear you already have everything the drawing shows, in more detail.

The two levels are drawn with deliberately different weight, and the reason is the same reason the wording is careful. Speech can hedge — "measured target 39,400 if it breaks" — but a bold line at 39,400 simply says target. So the trigger is solid, because it is a real price where the formation really does confirm; the measured target is faint, dashed and labelled, because it is a convention nobody here has tested. The visual weight is the disclaimer.

At most three formations are drawn at once, for the same reason the spoken readout describes one and counts the rest: five overlapping shapes hide the price they are describing.

Timeframes

Every tolerance in the detector is expressed in ATR — the instrument's own volatility — rather than in percent or in dollars, so nothing needs recalibrating when you change timeframe or switch from a $3 small cap to a $600 index fund. Measured across 1-hour, 4-hour, daily, 2-day and weekly bars on the same set of markets, the share of bars carrying an announcement stays between 8.1% and 9.2%, and the number of announcements per formation stays between 1.69 and 1.71. Those are the numbers you would want to be flat, and they are.

The one thing that is counted in bars rather than time is how long a formation may run: a shape must span at least 12 bars and at most 160. That is intentional — a "double top" whose two highs are two years apart is not a double top, it is two highs — but it does mean a formation is always sized relative to the chart you are on, never to the calendar. A 12-bar flag is an hour on a 5-minute chart and three months on a weekly one, and both are flags.

Detection runs once when a chart loads and is then cached, so it costs nothing as you navigate. On a 5,400-bar daily chart it takes about 20 milliseconds; on 328,000 intraday bars, about two seconds.

Heikin-Ashi

Formations are always read from standard candles, even while Heikin-Ashi is displayed — and the terminal tells you so when you switch HA on with formation description enabled.

The reason is the levels. A Heikin-Ashi close is an average of four prices, not a price anything ever traded at, so a neckline measured from one cannot be put into an order. The trigger and the measured target are exactly the numbers you might act on, so they are taken from real prices. Heikin-Ashi also smooths away the wicks that define a double top's two peaks, which means shapes that exist on an HA chart may not exist in the market and vice versa.

The practical consequence: with HA on, your spoken open/high/low/close are Heikin-Ashi values, but the formation levels beside them are not. Use HA to hear trend more clearly; trust the formation levels as real prices.

The same applies to the body and wick percentages, from both the arrow keys and the detail key: with HA on they describe the Heikin-Ashi candle you are looking at, not the raw bar underneath it. That matters more than it sounds, because a trending HA series routinely produces candles with no shadow at all on one side — the shaved look is the whole point of the smoothing — where the underlying bar has one. If those two ever disagree for you, the readout is wrong and worth reporting: they are meant to describe the same candle.

How many formations to expect

Roughly 5 formations per 100 bars, and about 8–9% of bars carry an announcement — one every twelve bars or so. That holds steady across 1-hour, 4-hour, daily, 2-day and weekly charts, because the detector measures everything in ATR.

That rate is an output, not a dial: it is what the detector happens to find, and nothing you can set changes how the shapes are identified. It is measured after every change to the feature, because both real defects found in it so far were properties of a rate rather than of any single sentence — one version announced each formation on exactly one bar and never again, and every unit test still passed.

Nested formations are normal, not a fault. A large inverse head and shoulders can genuinely contain two ascending triangles, in the same way a paragraph contains sentences. The larger shape is what the chart is making; the smaller ones are detail inside it. That is why the readout ranks by size and describes the biggest live one first.

Value zones

Value Deviation answers a different question: not where the swings are, but where value is, and where price has historically refused to go far from it.

It builds a rolling volume profile over the last N bars (the Profile window, 240 by default) and takes its point of control — the price where most volume traded — as value. Then it measures how far each bar strayed from that value in units of the value area's own width, and when price reverses at a distance, it marks that bar. A reversal below value marks a support zone; a reversal above it marks a resistance zone.

The mark's shape and colour carry how far from value the zone formed, across five tiers per side:

Each tier has its own pitch too, so the tier is audible as well as visible — the support tiers descend in pitch as they deepen, the resistance tiers rise.

Show tiers from is the density control, and it defaults to 2. Tier 1 is a reversal barely outside value — closer to noise than to a zone — and on a long view it is the bulk of the marks; left in, it turns the price pane into a continuous band of glyphs. Raise it to 3 or 4 on a weekly chart or a wide zoom to leave only the deep stretches, or drop it to 1 to see everything the analyzer found. It hides the glyph only: the Deviation Tier component and the spoken detail still report every tier, so navigating to a bar still tells you it was a tier 1, and nothing you could act on is hidden from speech.

Require a momentum turn as well is on by default: a zone is only marked when the indicator's own internal WaveTrend oscillator is turning the same way, which gives fewer and better-confirmed zones. Turn it off for more zones and more noise. The indicator carries its own oscillator maths internally, so you do not have to add Cipher B to the chart to get the benefit.

This one, too, is descriptive rather than a buy/sell signal. What it tells you is "price came this far from value and turned here" — which is what makes a level a level. Its own testing is worth stating plainly: the mean reversion it rests on is real but small in equities at a roughly five-day horizon, and it did not replicate as a tradable edge in crypto beyond Bitcoin. Read it as a map of where zones formed, not as a promise about where price goes next.

The Profile window adapts down if the chart has not loaded enough history — a window can never exceed a third of the loaded bars, because otherwise a fresh 200-bar chart would leave every component empty and read as broken. More loaded history genuinely helps this indicator, so zoom out and let it fetch before judging it.

The respect report

Every chart has more lines on it than matter. The respect report tells you which ones this market actually reacts to — measured, not assumed.

Open it with Alt+R, or with the Levels button on the toolbar's first row. It measures over the loaded history and presents two tabs: Levels near price (support and resistance levels from the indicators you have on the chart, prior period highs and lows, round numbers) and Moving-average ranking (a standard set of periods — 10, 20, 21, 50, 89, 100, 200 — including higher-timeframe projections).

Each row is a table row with proper headers: the line, its current price, its hold rate, how many touches it has had, the median reaction size in ATR, how often it held as support versus as resistance, when it was last touched, and how far away it is now in ATR.

The definitions matter, and the dialog states them under the table:

• A touch is counted when the bar's range — wick included — reaches within a tolerance of the line, and only if enough bars have passed since the last counted touch that this is a separate event rather than the same one over again.
• It held if price then moved a minimum distance away within a set window without closing through the line by more than the break tolerance.
• Wicks through and straight back count as holds. That is a sweep — the level working, not failing. Measuring the reaction from the wick instead of the close would make a sweep and a genuine breakdown look identical, which was the first version's bug.

Only lines with at least N touches is ticked by default. Leave it ticked: a line with two touches and a 100% hold rate is not a reliable line, it is two coincidences, and rows below the threshold are marked "(thin sample)" when you untick it. Re-measure recomputes after you have loaded more history or changed indicators, and Speak summary reads the top of the ranking aloud.

The asset dossier (Alt+I)

Alt+I — I for Instrument — opens a report on whatever is loaded on the active chart. There is no symbol picker inside it, deliberately: a second selection would drift out of sync with the chart you are reading, and "the dossier is describing a different asset than the chart" is a bug nobody would catch by ear. Choose market, provider and symbol on the toolbar, load the chart, press Alt+I.

Everything is looked up live at the moment you open it. The asset class comes from the market you loaded from, not from the ticker — "ETH" is a coin on Bitstamp and could be an equity ticker elsewhere, and guessing from the symbol would produce a confident report about the wrong kind of thing.

The tabs are questions, not sources. Tabs labelled "CoinGecko", "GitHub" and "SEC" would push the synthesis back onto you — four tabs and four half-answers to hold in your head to decide one thing. So you get:

Every individual field still names its own source. The chart read never needs a network and is always the first tab, so the dossier is useful even with every remote source down.

When there is nothing to show

This is the case the feature exists to handle well, so every row carries one of four states and a blank row is treated as a defect:

No data and unavailable are never merged: the first is a finding, the second is a reason to retry. For an unlisted token the dossier says so outright — "No listing found. For a very new token this is expected, and it means nothing here can be verified — treat every claim about it as unchecked." That is the most informative thing the screen can say about a brand-new token, and it would be lost in a blank panel.

What makes it more than a price page

Price, market cap and rank are the front page, and repeating them adds nothing. Two things are not on any price page:

Is anyone still building it? The dossier queries GitHub directly on the repositories a project lists, and when everything listed looks stale it sweeps the owning organisation for its most recently pushed work. This matters more than it sounds. Measured while building the feature: CoinGecko reported Kaspa at zero commits in four weeks, because it tracks a repository that was superseded — while the actual project repository had been pushed that same day. Reading the aggregator alone shows one of the most active projects in the market as abandoned. When the fallback sweep is used it is labelled as what it is: activity in the same organisation, not necessarily the flagship.

Does it disclose anything? Website, whitepaper, public source, block explorer — present or MISSING. The absence is the measurement. No whitepaper or no public source is the loudest cheap signal there is.

Plus supply and dilution (FDV against market cap, circulating share of maximum, uncapped issuance) and turnover, flagged at both ends — too little and you cannot exit, too much is a wash-trading tell.

The checks are not a score

Eleven checks for crypto, each one comparison over a value already on the screen, each shown with its own reasoning, and never summed. A single number would read as a rating.

None of these thresholds has been tested against forward returns. They are conventional red flags and the dossier says so. Testing them properly needs point-in-time snapshots of a universe that still contains the tokens that died, and today's listings by construction do not — which is why this project records the crypto universe daily. The standing prediction, written down before any of it was built: this works as a veto, not as a timing signal. It should avoid losses; it should not pick winners.

Reading it by ear

The headline is a live region, so it is spoken when the dossier opens. Speak summary repeats it — useful when the summary is long, or when you have arrowed away and want it again. Tab moves between the tab strip and the panel; arrow keys move along the tabs.

Limits worth knowing

• CoinGecko rate-limits the free tier. The dossier makes one coin call plus up to four GitHub calls per open.
• GitHub allows 60 unauthenticated requests an hour. The organisation sweep only fires when the listed repositories look stale, to conserve that budget.
• SEC EDGAR covers US filers only. ETFs, index vehicles and non-US listings are not filers, and the dossier says so rather than showing an error.

Quick trade from the chart

You can size and place a trade without leaving the chart or opening the trading dashboard.

What the percentage means, and why the stop comes first

The percentage can mean one of two things, and Settings → General → "Quick trade: the risk percentage means" chooses which.

• The position's value (the default). 1% of a 100,000 account is a 1,000 position, whatever the stop. This is how an exchange ticket behaves; the stop is purely protective, and what you stand to lose depends on where you put it.
• What I lose if the stop is hit. The "1% rule": 1,000 is the cash you are willing to lose, and the quantity is that budget divided by the distance to the stop. Entry 43,700 with a stop at 42,100 is 1,600 of risk per unit, so 1,000 buys 0.625 BTC. A tight stop makes this a large position, which is the half that surprises people.

Either way the terminal will not place until you have set a stop, because without one the size is either a guess or a number you cannot check by ear. Arming speaks the budget and asks for the stop:

"Armed 1 percent, $1,000.00 at risk. Move to the bar for your stop and press control alt shift X. Escape cancels."

and setting the stop speaks the whole calculation, naming both the number you chose and the one that follows from it:

"Long 0.625 BTC, entry 43,700, stop 42,100. Position value 27,312.50. If the stop is hit you lose 1,000.00. Shift enter for a limit at the cursor, control enter for market. Escape cancels."

That sum is what a sighted trader does in a position-size calculator before every trade. Not having to leave the chart to do it is the point of the feature. Quantities always carry the instrument's own unit — "0.625 BTC", never "0.625 units". If the position would cost more than the cash in the account you hear a caution, not a refusal, because margin accounts hold positions larger than their cash.

The rest of the behaviour

Direction is inferred, never asked. A stop below the current price can only be protecting a long; above it, a short. There is exactly one right answer, so the terminal does not ask.

You are told you are armed on every bar you move to — "Armed 1 percent, stop needed." then "Armed 1 percent, ready." It is short because you hear it constantly, and unconditional because forgetting you are armed and then pressing Enter for some other reason is the one way this feature could cost you money. Placing says only "Market buy sent."; the fill that follows is the sentence that matters, and it is never talked over.

Escape always cancels, and it reaches an armed trade before it reaches a half-placed drawing — the armed trade is the one with consequences.

The stop is always sent with the entry. Your size was derived from the stop distance, so an entry placed without it would have a quantity justified by protection that does not exist. On brokers with native bracket support (Alpaca) all three legs go as one order; on the others the terminal verifies afterwards that something protective actually reached the exchange, and says so loudly if not.

Limits and markets price differently. Ctrl+Enter uses the live price. Shift+Enter uses the bar under your cursor — and re-derives the size from there, because if you have moved a long way since setting the stop, the stop distance has changed and so has the correct quantity.

Drawing tools

Drawing tools place reference lines and shapes that then become audible as you navigate — the cursor will announce when it crosses a trendline you have set. Every tool uses sequential anchoring, which is what makes them fully keyboard- accessible: you do not hold and drag, and you do not press Enter. Instead you navigate the cursor to a point and press the tool's own shortcut to drop an anchor there; navigate to the next point and press the same shortcut again to set the next; and so on until the shape is complete. Escape cancels a drawing in progress.

Take a trendline (Alt+Shift+T) as the model. Arrow to the first point and press Alt+Shift+T — speech confirms "Trend line: first point at {price}, {time}. Navigate to the second point and press the shortcut again." Arrow to the second point, press it once more, and the line completes: "Trend line placed, second point at {price}, from {price}." Three-point tools — Fibonacci extension, Andrews' pitchfork, Risk/Reward — simply take a third press.

Every tool names its own points rather than numbering them, because the number is the one thing that cannot help you decide where to put it. The risk/reward tool (Alt+Shift+P) asks for the entry, then the stop loss, then the take profit; the measure tool (Alt+Shift+M) asks for the start and the end of the move. Tools that work something out say it on completion: the measure tool adds its distance, percentage and bar count, and the risk/reward tool its ratio — "Risk to reward, 1 to 3.00."

Two chords are easy to confuse and are worth learning apart: Alt+Shift+R is the rectangle, Alt+Shift+P is risk/reward, and Alt+Shift+M is the measure tool, which takes two points and not three. Single-anchor tools — a horizontal price line, a vertical time marker, a text label, an anchored VWAP — finish on the first press, placing immediately at the cursor.

On a touchscreen with no keyboard, arm the tool from the chart menu's Drawing Tools, then use the touch toolbar's Place drawing point button in place of the shortcut: move the cursor with the bar buttons (or tap a bar) and press Place drawing point once per anchor. Re-pressing it advances through the same tool's anchors exactly as re-pressing the shortcut would, so a trend line, channel, or any multi-anchor tool completes entirely by touch.

The drawing chords are Alt+Shift+<letter> on every head — desktop, mobile and browser. A shortcuts profile saved before 2026-09-05 keeps the old Ctrl+Shift chords on the desktop, and the F1 help always shows the bindings live on your host.

The set spans Trendline (T), Horizontal and Vertical lines (H, V), Channel (C), Fibonacci retracement and extension (F, E), Text label (L), Rectangle (R), Measure (M), Andrews' pitchfork (A), Gann fan and box (G, B), Angle (J), Risk/Reward (P), and anchored VWAP (W); Alt+D opens a panel to review and delete what you have placed. A few behave specially. A Fibonacci retracement, anchored from a swing low to a swing high, lays down the standard 23.6, 38.2, 50, 61.8 and 78.6 percent levels, each audible as the cursor crosses it. An anchored VWAP behaves like a moving-average overlay you can focus and whose crossings you can jump between. And the Risk/Reward tool, after you set its entry and stop anchors, speaks the resulting risk and then asks for the target, announcing the full reward-to-risk ratio once you set it — the same measuring workflow described in the Trading chapter.

Arrowing along a drawing

Focus a drawing with Page Up or Page Down and the bars read differently from an indicator's. The switch names it once — "Trend line 2." — and each bar then says the drawing's value first, where you are on it second, and which side of it price is on third, with the middle and last parts left out whenever there is nothing to say: "150.50, price below." between the anchors, "170.50, at end anchor, price above." on an anchor, "199.50, past end, price above." where the line is projected beyond the point you drew it to, and "price crossed above" on the bar where the close changed sides. Where the drawing has no value at all, the position leads with a bar count — "Before start, 20 bars." — because arrow keys move bars and a count tells you how far to press. A drawing with several parts (a Fibonacci retracement's levels, a rectangle's top and bottom) announces which one it is reading — "Fibonacci retracement 1. 7 components, reading 61.8%." — and Ctrl+Up and Ctrl+Down name each level once as you move between them; the bars themselves stay value-first. "Not yet calculated." means the drawing genuinely has no number at a bar it should — that is a fault worth reporting, not a place you are standing.

Moving an anchor afterwards, from the keyboard

Until version 2.6 an anchor you had already placed could be moved only by dragging its handle with a mouse, or by typing an exact price and date into Properties. Typing an exact value is a route, but it is not the same thing as "a little to the right", which is what the drag was for. Now it has a keyboard equivalent. Focus the drawing with Page Up or Page Down — or pick it in the Object Tree — and:

• Shift+Left and Shift+Right move the selected anchor one bar earlier or later. A bar, not a day: on a daily chart Friday's next bar is Monday, and a halt is stepped over the same way. Past the last bar the anchor projects into the chart's right margin, and the readback says so: "June 30, 3 bars past the last bar."
• Shift+Up and Shift+Down move its price by one percent of the price range you can currently see, so the step follows your zoom the way a drag would, and never by less than the last decimal the price is spoken with.
• Ctrl+Alt+Shift+G selects the next anchor. On a drawing you have just focused, the first press only tells you which anchor is selected; the next moves on, and it wraps.
• Ctrl+Alt+Shift+B snaps the selected anchor's price onto its bar's high, low, open or close — the nearest first, then the others in that order on repeated presses. It is the quick way to put a trend line's end exactly on a wick.
• Shift+F1, the context summary, names the selected anchor without moving it.

Hold a key and you hear a short tick while the anchor moves; one sentence follows when you stop — "End: 105.20 at June 15, 2026, 09:30. Trend line 2, anchor 2 of 2." — the value first, then which drawing and which anchor of how many, every time, because there is no status bar to glance at. A key that cannot act plays the boundary sound and says why once: the anchor is already at the first bar, a Fibonacci level has no date to move, no drawing is focused — or the chord was pressed where it cannot reach the chart: "The chart does not have focus. Control Alt Shift C returns to the chart." with focus on a toolbar button, and "Not while Properties is open. Escape closes it." under a dialog. The Object Tree is the exception, on purpose: it is where you pick a drawing, so the nudge works while the tree is open and you can move what you just selected without closing it first. A whole run of nudges is one Ctrl+Z. The same six actions are in the drawing's context menu (Shift+F10) for voice control, switch access and single-pointer use, and the menu stays open so "move later" can be activated several times.

Two platform notes. On Windows with more than one keyboard layout, Alt+Shift pressed and released on its own switches layout, so hold Alt+Shift and then press the arrow. On a Mac with VoiceOver's modifier set to Control+Option, every Ctrl+Alt+Shift chord in this application is a VoiceOver chord: set the modifier to Caps Lock in VoiceOver Utility, or pass a single key through with VO+Tab.

Taking a drawing back

Ctrl+Z undoes the last chart edit and Ctrl+Y redoes it, and both tell you what they just did rather than only that they did something — "Undone: trendline moved.", "Nothing to undo." The stack holds the last fifty edits, which is deep enough that "undo until it sounds right" works. It covers the two things that are genuinely painful to reconstruct by hand: edits to a drawing, and deleting a series with Delete. It is not a general application-wide undo — it will not take back an order, an alert, or a settings change, and nothing that leaves the chart is on it.

The context menu, from the keyboard

The Applications / Context Menu key, or Shift+F10 if your keyboard has no such key, opens the same right-click menu a mouse user gets. What you get depends on what is focused: a focused drawing gets the drawing's own menu — Delete, Duplicate, Properties — and anything else gets the chart-level menu, the one right-clicking empty chart space opens, carrying the bar under your cursor so options like "Play from here" act on the right bar. It is exact keyboard parity with right-click, so nothing in this application is mouse-only.

Naming your drawings

A drawing's name is what you hear when you move to it, what the object tree lists, and what an alert refers to. By default that is "Trendline 3", which tells you nothing. Open Properties (P) on a drawing and set Name to why you drew it — "Weekly resistance from the January high" — and that is what is announced from then on. Clearing the field restores the automatic name.

For a keyboard user moving between a dozen drawings by ear, this is the difference between a list and an inventory.

The volume profile

A volume profile is a different kind of view: instead of volume across time, it shows volume across price — a horizontal histogram revealing which price levels saw the most trading. When your focus is on a volume-profile series the Up and Down arrows change meaning: rather than stepping through components they move between price bins, Up to the next higher level and Down to the next lower, each announced with its price and the volume there. Left and Right still move the cursor through time as usual.

Two landmarks are worth listening for. The Point of Control — the single highest-volume price level — rings with a distinct square-wave tone and is announced "Point of Control" when you land on it. The Value Area, the band of prices that accounts for roughly seventy percent of all volume, announces "Entering Value Area" as you move into it and "Exiting Value Area" as you leave — so you can feel the edges of where the market agreed on price.

There are eight profiles in the Add Indicator dialog's Profile category, and they are one idea crossed with another rather than eight different tools. The first choice is what is counted: a Volume Profile counts the volume traded at each price, and a Market Profile (TPO) counts how many time periods price spent at each level, so a level price lingered at ranks highly even on thin volume. The second choice is which bars, and there are four windows:

• Visible Range — whatever is on screen. It recomputes as you pan and zoom, so it always answers "what am I looking at".
• Fixed Range — the stretch you were viewing when you added it. It stays put while you navigate away, so you can compare current price against a reference you chose. Properties reads the range back to you as a sentence.
• Session — one trading day, the day of the last bar on screen. Pan into yesterday and you hear yesterday's session; this is how a market profile is conventionally read. Meant for intraday charts — on a daily chart a session is a single bar.
• Anchored — from the bar your cursor was on when you added it, running to the newest bar and growing as bars arrive. It is the anchored-VWAP idea applied to volume: where has trade concentrated since that event? Properties tells you the anchor date, and to start from a different bar you add a new one while on it.

Keeping a visible and a fixed or anchored profile on the chart at once is the normal way to use them: one says what you are looking at, the other what you chose.

The heatmap

Alt+H toggles a volume heatmap over the price chart, shading each candle and time zone by how much trading happened there. Its real payoff is in sound: with the heatmap on, playback folds that intensity into the audio, so higher-volume bars play louder than quiet ones and a busy stretch of market is something you hear swell rather than something you have to look up. Press Alt+H again to turn it off.

The object tree

Once a chart has a few indicators and drawings on it, navigating to a particular one just to hide or remove it becomes a chore. The object tree, opened with Alt+O and announced as "Objects: chart object tree", is the manager's-eye view that solves that: a single tree of everything on the chart, laid out as the same panes → series → components hierarchy you move through on the chart itself, but as one list you can walk through without disturbing the chart.

It is a real tree, so it answers to the tree keys rather than to Tab: up and down arrows move between entries, right and left expand and collapse the entry you are on, and Home and End jump to the first and last. Tab is still how you reach an entry's Hide, Mute and Delete buttons, and Escape closes the panel. Each entry says whether it is expanded or collapsed as you land on it, so collapsing a pane and re-opening it sound different.

Each series in the tree reads itself the way you'd want — its name, how many components it has, and its current state — so you hear, for example, "RSI, one component, visible, audible, focused, collapsed" as you move onto it. From there you act on any object in place, without first navigating to it on the chart: each series and each component has a toggle to hide or show it (the same effect as H) and a toggle to mute or unmute it (the same as M), and a series carries a Remove control to take it off the chart entirely. Selecting a series in the tree also focuses it back on the chart, so the panel doubles as a jump-to-anything: find the indicator in the list, activate it, and your chart cursor is now on it. When the soundscape is getting crowded, this is the quickest place to find the one series cluttering it and silence or remove it. A button at the top also jumps straight to the Strategy manager.

AI, Narration, and the Journal

Three features sit between reading the chart yourself and acting on it: an AI analyst you can ask for a second opinion, an auto-narrator that watches a series and speaks up when something happens, and the Journal that quietly records everything the terminal has said so nothing scrolls past for good.

What speaks when

Everything the terminal says without being asked fits one table. Three questions decide each row: N picks what may speak, the Narration tab picks when, and the scope you played picks which of them. Where the words go — spoken here, or a system notification — is the subject of the Monitoring chapter, and it depends only on whether the event is about the chart in front of you.

What never speaks unprompted, whatever you flag: a continuous line's value (that is what the tones are for), your drawings (they read when you arrow onto them), a comparison overlay, and the unbounded accumulators — OBV, A/D, Force Index, ATR, standard deviation — which have no threshold that means anything. All three chords in this chapter use three modifiers and are the same on every head.

The AI technical analyst

Press Ctrl+Alt+Shift+A to open the AI Analyst. It gathers a snapshot of what you are looking at — the current symbol and timeframe, the most recent candles, a summary of every indicator you have on the chart, and, where the provider supports vision, an actual image of the chart — sends it to a large language model, and reads back a concise, plain-language technical analysis written for text-to-speech: trend direction, the key support and resistance levels, what momentum is doing, and a short-term outlook. Think of it as a second pair of eyes on demand — a narrative framing of the same structure you have been navigating bar by bar, in one paragraph you can take in at listening speed.

It is worth being clear about what it is and is not. It describes the chart; it does not advise you and it cannot place a trade — there is no button in it that touches your account. And because the answer is generated, it can be wrong: treat it as informed commentary to weigh against what you heard yourself navigating, not as a signal to act on. Used that way — as a sanity check on the read you already formed — it earns its place.

The natural way to use it is to set the scene first. Load the symbol and timeframe you care about, add the indicators you want it to consider, navigate enough to have your own opinion, and then press Ctrl+Alt+Shift+A — because it analyses the chart as it stands at that moment, the more you have set up, the more grounded its answer. A useful habit is to ask twice: get its read on the daily, switch the timeframe to the hourly (or add an indicator you suspect matters), and run it again to hear how the framing changes. A typical reply sounds like this:

"Bitcoin, four-hour. The trend is up but stretching — price is roughly eight percent above the fifty-period moving average and RSI is at seventy-three, in overbought. The nearest support is the prior breakout around sixty-one thousand five hundred; resistance is the recent high near sixty-four thousand two hundred. Momentum is still positive but the MACD histogram is shrinking, which often precedes a pause. Short-term outlook: constructive but extended — a pullback toward the moving average would be normal and would not break the uptrend."

It needs a key. Add one for at least one provider in the API key manager (Alt+K); the terminal tries the providers it knows in order — Claude, then OpenAI, then Ollama — and uses the first you have configured. If none is set up, the Analyst tells you so rather than failing silently. The choice of provider is partly a privacy choice. Claude and OpenAI are cloud services and are vision-capable, so they get the richest input — but your chart snapshot, image included, leaves your machine to reach them. Ollama runs a model locally on your own computer, so nothing leaves the device at all; it is the right pick when you would rather your data stay home, at the cost of running a smaller model and installing it yourself. Either way, remember that a cloud request both shares your chart data and usually costs a small amount per call, so it is a deliberate action, not something to lean on every bar.

Auto-narration and live announcements

Where the AI Analyst is something you ask, auto-narration is something you switch on and forget. Press N — the third switch on a chart object, beside H for hide and M for mute — and from then on the terminal watches that indicator and speaks new events as they occur on live bar closes — a fresh signal firing, or the oscillator entering or leaving an overbought or oversold zone. You will hear short, plain announcements as they happen: "RSI overbought", "MACD bullish crossover", "Stochastic leaving oversold". It announces only what happens after you switch it on; it does not replay the history you already navigated. And because it is per-series, you can leave it running on the one indicator you care about and not be interrupted by the rest of the chart — set it on your RSI, say, and get on with reading price while the terminal keeps half an ear on momentum for you. Toggling it announces the new state, "Narrating" or "Narration off", so you always know whether it is listening. Hiding a series or a component (H), or muting it (M), takes it out of narration for as long as it is hidden or muted — on bar closes and in playback alike — without touching the N flag, so showing it again brings its narration straight back. Flag a series with nothing to narrate and the terminal says so instead of promising: "{Series} has no signals to narrate. Press 0 to add a reference level and its crossings will speak." (The former Ctrl+Alt+Shift+N alias is gone; N is the switch, and it needs the chart focused.)

N follows your cursor, exactly like H and M. With the cursor on a series it switches the whole series; with the cursor on a component — Ctrl+Up and Ctrl+Down move between them — it picks out that one component. That matters on an indicator like Cipher B, which prints eleven components: switching the series on and leaving it there gives you all of them, and pressing N on the divergence line narrows narration to just that. Press N on it again and you are back to the whole series. So the rule is: no component picked out means all of them, and the series switch is always the master — a component flagged on a series that is not narrating says nothing yet, and the confirmation tells you so rather than leaving you waiting. The confirmation on a component names the component and nothing else — "Triple Confluence Buy, narrating" — because the cursor is already on it and the series was named when you arrived. When you want everything back to a known state, Ctrl+Alt+Shift+O switches narration off on every series and clears every component selection, the way Ctrl+Alt+Shift+K and U undo every hide and every mute.

Where these words land in the sentence. Moving onto a series or component that is switched off says so first — "Hidden and muted. Cipher B. 11 components…" — and that it is narrating last: "…64,900. Narrating." The two halves are not the same kind of fact. Hidden and muted explain a silence, and anything that interrupts an utterance takes its end, so the half you must not lose goes in front; narrating is an addition, everything else about the reading is normal, and it is the least urgent thing in the phrase. Both flags are spoken when both apply, because they are cleared by different keys and a readout naming one of them is a readout that guarantees a second wrong guess.

What an overlay narrates, and when. Flag a moving average with N and what you get is its crosses, on the bar close: "Price crossed above EMA 9 at 64,900." An overlay gets crosses and nothing else — no "support broken", no "tested twice" — because those belong to a level, which has a side and can cease to exist, and a moving average has neither. Cipher SR's pivots and Spider Lines' fibonacci EMAs are declared levels and do get the fuller vocabulary. And none of it is spoken during playback; see the playback section above for why.

What an oscillator narrates, and when. Its levels — the thresholds the indicator was built around, which it declares for itself and which you can see as dashed lines in its pane. Crossing one is spoken on the bar close: "Stochastic 14: crossed above overbought, 80.", "Rate of Change 9: crossed above zero." Move a level with the Properties dialog, or drop your own with 0, and the new value is what gets narrated — the sentence follows the line, not a number baked into the terminal. A handful of indicators have wording written specially for them and use that instead, so RSI says "RSI overbought" rather than the generic form, and Cipher B says "Anchor wave oversold"; you never hear both for one crossing.

Every indicator, and what it will say. Six routes, and an indicator uses whichever ones it has:

Four kinds of thing stay silent however you flag them, and it is deliberate: your drawings (they answer to the drawing keys and read themselves when you arrow onto them), the candles, volume and the profile surfaces (the new-bar announcement is the candle's narration), the comparison overlays (another symbol's price, drawn here), and the unbounded accumulators — OBV, A/D line, Force Index, standard deviation, ATR, Ulcer, historical volatility. Those last ones have no threshold that means anything: an ATR of 400 is enormous on one asset and noise on another, so there is nothing to cross and inventing a number to cross would be worse than the silence.

What a signal sounds like. A narrated signal is introduced by the component that fired it, never by the series: "WaveTrend Cross Bull: Wave cross up 12." You chose which series narrate, so the series name tells you nothing you did not already know; which of Cipher B's eleven markers just printed is the fact you are waiting for. Where a signal's own wording already says which component it is — "Bullish divergence", "Triple confluence buy, strong confirmation" — nothing is put in front of it. The same rule holds during playback.

The date, only when it changes. Arrowing along an hour chart reads the time alone — "14:00. Close 64,900.", "15:00. …" — and names the day when the bars you are reading cross into a new one: "September 6, 00:00. …". It follows the reading rather than the chart, so arrowing backwards across midnight names the day you have just come back into, and so does a jump that lands in a different one. On a daily chart or coarser the date is the timestamp; you no longer hear "00:00" after it, which was the same on every bar a daily chart has. If you would rather have the date in front of every reading, Settings (F12) → Speech → Speak the date on every bar turns it on; boundaries only is the default. The two "Speak values as" orders that name a timestamp explicitly — time only, date only — are unaffected: choosing one has already answered the question.

Everything about one bar arrives as one sentence. When a bar closes you hear the candle and then whatever your indicators made of it, in one breath: "Close 64,905 at 14:32, Bullish engulfing. New bar: Open 64,910. Triple confluence buy, strong confirmation. RSI 14: RSI overbought." The closing bar is named by its own stamp — the time of day on an intraday chart, the date on a daily or weekly one — so a run of them in the journal reads back as a record rather than a column of prices. Up to five clauses, ordered by consequence — a level breaking first, then an indicator's own signals, then crossings, then tests, then approaches, and last the oscillator commentary, which is the most frequent thing the terminal can say and the least worth leading with. Anything beyond five is dropped from the bottom of that order rather than from whatever arrived last.

Two switches sit above all of it, both on the Narration tab of Settings (F12). "Narrate signals on bar close" is the master: N chooses which series and components speak, that switch decides whether any of them do, and turning it off is how you get an hour of quiet without un-flagging every indicator and having to remember what you had flagged. "Narrate during playback" decides whether those same flagged series speak while the chart is playing, covered back in the playback section — and what you hear there is scoped to what you played, so Shift+Space on one series narrates that series and nothing else on the chart. One edge worth knowing: switching the master back on can replay a signal up to twenty bars old, because a pivot takes that long to confirm; it is bounded, and a twenty-bar-old signal is usually still worth hearing.

Auto-narration is one of a small family of "let the terminal keep you posted" features worth knowing together. The rolling new-bar announcement — the "Close … New bar …" you met when you first loaded a market — is the always-on heartbeat of the live candle, and you can turn it on or off under Settings (F12), Narration, with "Announce new bars". The detailed point analysis, Alt+Shift+D, is the on-demand deep read of whichever bar you are sitting on — candle values, patterns, every indicator, every signal, in one keystroke, covered back in the chart chapter. And the context summary, Shift+F1 for a quick "symbol, provider, timeframe" and Ctrl+Alt+Shift+C for the fuller picture, tells you where you are at any moment. Between them you can run as hands-off or as hands-on as you like: narration and new-bar announcements stream the live market to you, while Alt+Shift+D and the AI Analyst are there the moment you want to stop and look hard at something.

The Journal

The Journal, opened with Ctrl+Alt+Shift+J, is the terminal's memory of the session. Everything it has spoken or alerted — ordinary speech, alerts, strategy setups, errors, and backtest results — is written here, newest at the bottom, up to a couple of thousand entries. It is the primary way to recover something that went by in speech while you were concentrating elsewhere: a fill you half-heard, an alert that fired mid-thought, the exact wording of an error.

The view is a plain monospace text area you can Tab into to read or copy any line, with filter buttons — All, Speech, Alerts, Setups, Errors, Backtests — to narrow it to just the kind of entry you are after, and a "Copy visible" button to lift the current selection out to paste elsewhere. Automated-strategy setups land here in full, with their rationale spelled out — side, score, stop price, first target, reward-to-risk, and the notes on where the stop was placed — so the Journal doubles as the record of what your strategies proposed and why. When in doubt about what the terminal just told you, this is where you go to read it back at your own pace.

Trading

Everything to do with money — placing orders, attaching protective exits, watching positions and fills, and reading the live order book — runs through the trading dashboard, which you open with Alt+T. Your screen reader announces it as "Trading Dashboard". It gathers three things in one place: the market you are trading and its environment, an order ticket, and a row of account tabs — Positions, Balances, Orders, History, and Book. This chapter assumes you already know what market, limit, stop, and trailing orders are, and concentrates on how you express and hear those decisions here, and on what the terminal does and does not do on your behalf.

Practise first: paper trading

Before risking a cent, turn on paper trading mode and learn the whole workflow against simulated money. Open Settings with F12, and on the General tab tick "Paper trading" and press Save; the terminal confirms "Paper trading enabled" and a PAPER badge appears in the status bar (announced as "Paper trading enabled"). From then on, every order you place — on any chart, with any provider, even a data-only one — is routed to a built-in simulator instead of a real exchange. Fills are driven by real live prices: a market order fills at the current price, and a stop, target, or trailing order fills the moment live price action actually crosses it, less a simulated 0.04% taker fee. You start with 100,000 USDT, the account persists between sessions, and a "Reset paper account" button on the same settings tab wipes it back to the starting balance whenever you want a clean slate. It asks you first. The first press arms a confirmation and says what is at stake — the balance, every open position and working order, and the whole trade history, with no undo — then offers "Confirm reset" and "Cancel", with focus on Confirm. Escape at that point backs out of the question rather than closing Settings; press it again to close. Whichever way you answer, focus returns to the button you pressed. While paper mode is on the dashboard shows the environment as "Paper (simulated)" with a paper banner, and the red live-funds banner is suppressed. Everything described in the rest of this chapter behaves identically in paper and live — so practise here until the spoken feedback is second nature, then switch a real key in.

Your trades keep running when you look elsewhere. A chart you have an open position or a resting order on is watched in the background, so a stop fires and a limit fills while you are on a different tab entirely, and the position's profit and loss keeps moving rather than sitting still until you go back. This works even after you close that chart's tab, and after you close and reopen the app — the terminal remembers which chart each position was opened on. You do not need to switch background monitoring on for this; that setting governs watching charts you merely have open, while a chart with money on it is watched either way.

What the paper account will and will not do. It settles like a spot exchange out of a single cash balance, so it will refuse to sell an asset you do not hold and refuse a buy you cannot afford — telling you by how much in both cases. Brackets, one-cancels-other pairs, trailing stops and trailing take-profits are all fully simulated.

Short selling is available again, at 1x, with collateral. It was withdrawn in 2.2.0 because it was offered and did nothing; it returned in 2.3.0 modelled properly. Selling something you do not own means somebody lent it to you and you owe it back at whatever it ends up costing, so the account locks the sale proceeds and an equal amount of margin beside them — both appear under Locked on the Balances tab, and neither is spendable while the short is open. You are told what the short cost to open, what it is worth now, and the liquidation price: the level at which the collateral no longer covers the position and the simulator closes it for you. If you try to open a short the collateral will not cover, it is refused with the shortfall named rather than accepted and quietly unwound later.

Leverage is still withheld. The margin requirement is 1x — a dollar of collateral per dollar of exposure — so there is no borrowing beyond your own balance, and the leverage selector does not appear on the paper ticket at all. It returns when multi-x margin is properly modelled, and not before. The Cross / Isolated margin choice is live in paper, though: an isolated short is liquidated against its own collateral, a cross short against the pooled collateral of every cross position plus free cash, so you can rehearse the difference.

On the hosted web terminal (the logged-in, multi-user build) this choice is made for you: paper trading is always on and cannot be switched off, so pressing Alt+T always opens a paper dashboard and you can never place a real-money order from the browser. Real trading with your own broker keys is a desktop-app feature. (If you tried this on an earlier build and got "provider does not support trading," that was the bug this behaviour fixes — the web providers are data-only, and orders now correctly route to the paper simulator.)

Your paper account is one account, however many tabs you have open. Open the terminal in three browser tabs and all three are looking at the same balance, the same positions and the same resting orders; a trade placed in one appears in the others. This is worth stating because it was not always true: until August 2026 each tab kept its own copy of the account and wrote the whole thing back to disk on every change, so the last tab to save silently erased whatever the others had done. If you ever found a position you were sure you had opened simply missing, that is very likely what happened, and it can no longer happen.

Resting orders are watched from every chart you have open, not only the tab that placed them. A stop you set in one tab still fills while you are working in another.

Paper or live — check this first

When you are not in paper mode, your environment follows the API key profile you chose. The terminal names it in four places, because for a screen-reader user a red banner is not a warning:

• The status bar carries a PAPER badge or a LIVE badge ("Live trading — real money"). The live one exists because an absence is not a signal you can navigate to.
• The dashboard speaks the account when it opens: "{Provider}: LIVE account '{nickname}'. Orders here are real money." or "…paper account '{nickname}' — a practice environment, no real money."
• The "API key for this order" dropdown at the top of the dashboard is the choice of record: the key you pick is the key that signs. Its options read "{nickname} — Live, real money" or "{nickname} — Paper, practice environment", and changing it reloads that account's balances and positions. Anything not explicitly marked Paper is treated as live.
• The banner across the ticket reads "⚠ LIVE TRADING — Real funds at risk (key: {nickname})", for anyone looking.

A Paper key only means paper on a venue that has a practice environment. Alpaca, Binance, Gemini, Oanda and Tradier do: a Paper profile there points at the sandbox or testnet. Bitstamp, Coinbase, Interactive Brokers, Kraken, Kraken Futures (its demo was withdrawn on 2026-07-14), MEXC and Schwab have none, so a key marked Paper on those would sign against real money while everything on screen said paper. The terminal refuses such an order instead, at the one point every order passes through: "{provider} has no practice venue and the key '{nickname}' is marked Paper, so this order would be real. Mark that key Live to trade real money, or turn on Paper trading in Settings." Build the habit of confirming the environment out loud before a session — it is the single most consequential setting on the screen.

Placing an order

The order ticket lives in the dashboard's "Place Order" panel and reshapes itself to the order you are building, so you only ever Tab past fields that apply. You choose a side with the big round green "BUY" and red "SELL" buttons — they are a toggle, so exactly one is active and your screen reader reports which is pressed — then enter a "Quantity" and pick a "Type".

The Type list holds the full set: Market, Limit, Stop-Market, Stop-Limit, Take-Profit-Market, and Take-Profit-Limit. The fields that appear depend on it. The stop and take-profit types reveal a "Trigger Price" — the level at which the order activates. The limit-style types (Limit, Stop-Limit, Take-Profit-Limit) reveal a "Limit Price" and, with it, a "Time in force" choice of GTC, IOC, or FOK and a "Post-only (maker)" checkbox. A plain market order fills at the prevailing price and needs neither.

On a market or limit entry you also get the protective and sizing controls described in the next two sections — Stop Loss, Take Profit, trailing exits, and a risk sizer. And if your provider supports margin or futures, the ticket adds a "Margin" choice (Cross or Isolated), a "Leverage" multiplier, a "Position side" (One-way, Long, or Short, for hedge accounts), and a "Reduce-only" checkbox. Controls the active provider does not support are simply not shown — so the same ticket is lean on a spot exchange and full on a futures one.

Two helpers worth knowing before you submit. The Size button next to "Risk % of balance" does position-sizing for you: enter a risk percentage and a stop, and it sets the quantity so that being stopped out costs you that share of your balance, announcing "Sized {quantity} from {n} percent risk." And a double-tap guard ignores an identical order resubmitted within thirty seconds, so a stray second press will not double your size.

Protecting a trade: stops, targets, and trailing exits

You protect a trade at the moment you enter it. On a market or limit entry the ticket offers an optional "Stop Loss" and "Take Profit", both entered as absolute price levels — buy at 42,500 and you might type 42,180 to stop out and 43,400 to bank profit. Leave either blank to skip it.

When you would rather your exit follow the trade, use a trailing stop or a trailing take-profit (shown when the provider supports trailing). The "Trailing stop" selector lets you trail by Percent or by Amount; enter a distance and the terminal keeps a stop that ratchet toward the trade as price moves your way and never loosens, firing only on a reversal of that distance. The "Trailing take-profit" works the same way but adds an optional "TP activation price": it stays dormant until price reaches that level, then arms and trails from there, so it locks in profit only after a target is reached. In paper mode you can watch both work tick by tick.

How the protection actually reaches the exchange differs by broker, and the terminal is honest about each:

• Binance attaches stop-loss and take-profit natively with the entry (and supports true OCO pairs on spot).
• Tradier places the entry and its exits as ONE exchange-native advanced order (OTO with one protective leg, OTOCO with both) — the linkage lives on Tradier's servers, so your protection exists even if the terminal dies the moment after submission.
• Schwab does the same with an order tree: the entry triggers the exit (or an either-or pair of exits), enforced server-side, good-till-cancelled.
• Kraken can attach only ONE protective order per entry. If you set both a stop and a target, the STOP is attached — safety over profit — and the terminal tells you out loud that the take-profit was not placed.
• On the remaining brokers the terminal submits the entry, then the protection, then about two seconds later verifies the protection actually landed — and if it cannot find it, warns you, interrupting: "Warning: no stop loss or take profit found on the exchange for {symbol}. The position may be unprotected — verify your open orders." Treat that as a call to action.

Your resting protective legs are visible: on Tradier and Schwab the Orders tab lists each leg of a bracket individually — entry, stop, and target, each with its real trigger price — so you can hear that your protection is in place. To move a stop or target on an open position afterwards, use the inline editors on the Positions tab: each Stop loss and Take profit cell is a button that becomes a field; type the new level, press Enter, and it is checked against the current price before anything is sent. A resting order's own levels have no editor; cancel it on the Orders tab and place a new one.

The live order review

Paper orders submit the instant you activate the submit button. Live orders do not — they get a spoken safety review first. When you submit on a live profile, the terminal speaks a one-line summary — "Confirm: {side} {qty} {symbol}, {type}. Stop {price}. Target {price}. Estimated cost {amount}, fee {amount}. Confirm or cancel." — and replaces the submit button with Confirm and Cancel. The review also carries two warn-only risk checks. On a leveraged entry it estimates the liquidation distance and compares it with your stop: if the exchange could close the trade before your stop fires you hear a plain warning to reduce leverage, and if the buffer is under two stop-distances you hear a caution. And if the order stacks onto open positions in the same sector — Bitcoin and Ether are one bet, gold and silver are one bet, SPY and QQQ are one bet — a sector note reminds you that correlated exposure counts toward one 2-percent-per-sector risk budget. Neither ever blocks the order; the terminal informs, you decide. (Paper fills speak the sector note too, so the habit forms where the money is pretend.) Nothing reaches the exchange until you activate Confirm; Cancel backs out with "Order canceled before submit." It is the deliberate pause that a real-money order deserves, and it is on automatically whenever you are live. (The live-versus-paper decision is re-checked at the moment you submit, so flipping paper mode off in Settings with the dashboard open never lets a real-money order skip its review.)

Beyond that one-time check at submit, the terminal keeps watching every open leveraged position. If the market drifts close to a position's liquidation price while you hold it, you hear an interrupting spoken warning with an alert earcon — once per approach, naming the symbol and how far it is from liquidation — so a slow slide toward a forced close announces itself instead of surprising you. Positions with no liquidation price (spot holdings) are not monitored.

Hearing your fills

Order outcomes are the one kind of feedback the terminal will never let you miss. Whatever else is happening — sonification muted, a playback running, speech mid- sentence — an order event plays a short earcon immediately and then speaks over whatever was being said. The announcements are plain, quantity-and-price first, and on a closing fill they add the realized result:

• Placed: "Order placed. {Buy|Sell} {qty} {symbol}, {type}."
• A complete fill: "Order filled. Bought {qty} {symbol} at {price}." — and on a close, "… Profit {amount}." or "… Loss {amount}."
• A partial fill: "Partial fill. … {remaining} remaining."
• A fixed stop or target: "Stop loss hit. Sold {qty} {symbol} at {price}. Loss {amount}." / "Take profit hit. … Profit {amount}."
• A trailing exit: "Trailing stop hit. …" / "Trailing take profit hit. …", with the same price and profit/loss.
• A cancel: "Order canceled. {symbol}."
• A refusal: "Order rejected for {symbol}," followed by the reason — "Insufficient paper balance: that position needs 51,970.00 USDT and the account holds 174.91.", "No price available for BTC/USD: open its chart, or check the venue is reachable.", "The position was already closed." Closing a position from the Positions tab announces the same way. Before August 2026 the reason was computed and then discarded, so a refusal arrived as three words with nothing to act on; the half that told you what to do was the half being dropped.

Order IDs are never spoken — they are meaningless to the ear. Every one of these events is also written to the Journal (Ctrl+Alt+Shift+J), so a fill that goes by while you are concentrating elsewhere can be read back afterward. (Realized profit and loss are spoken on every paper close; on a live exchange they appear when the exchange reports them.)

A timing note for two brokers: most brokers (and the paper simulator) push order events instantly, so announcements are immediate. Tradier now has a live account-event stream too; if it can't connect, and always on Schwab (whose push channel isn't implemented yet), the terminal instead watches each order you place by polling the broker — so fills, stops, and cancels there still announce, typically within a few seconds (up to about half a minute for an order that has been resting a long time). One thing polling cannot see: protective legs the broker attaches server-side get their own order ids, so if one of those fires on Schwab you'll find it in the Orders/History tabs rather than hearing it named.

Positions, orders, balances, and history

The right-hand side of the dashboard carries five tabs — Positions, Balances, Orders, History, and Book — and switching to one speaks its name and count, like "Positions, 2".

• Positions lists each open position as the instrument as held ("BTCUSDT isolated 1x"), its direction spelled out, the exchange, quantity, average price, value, unrealized profit or loss, stop loss and take profit (each an inline editor), and liquidation price. Two buttons close it, each named for its order type: Close position flattens at market, Close at limit opens a price field first. A spot venue with no positions says why the list is empty rather than showing nothing.
• Orders lists your working, not-yet-filled orders — side, type, quantity, price, status — each with a Cancel button to pull a resting limit or protective stop.
• History is your fill log, newest first: time, symbol, side, quantity, price, realized profit or loss, and fee — the place to review how a session actually went.
• Balances shows, for every asset you hold: the free and locked quantities, what the holding is worth in the account's quote currency, its share of the account as a percentage, and the day's change. Above the table a summary line gives the portfolio total. An asset whose price cannot be fetched is shown as unpriced and the total says explicitly that it is partial — a price that could not be read is never silently counted as zero. Money locked as short collateral appears here under Locked (see paper trading, above).
• Book is a five-level snapshot of the bids and asks with the spread, refreshed every two seconds and read on demand by Tab; it never announces on its own. The full twenty-level order book is Alt+B.

Getting funds onto a venue: deposit addresses

Some venues can hand you a deposit address for an asset, and when the terminal can do that for the provider you are on, a Deposit button appears in the toolbar. Choose the asset and the network and the address comes back in a read-only text field with a Copy button beside it. That is deliberate, and it is the whole point of the dialog: a read-only field is walkable by your screen reader's review cursor and by a braille display character by character, so you can verify an address the same way a sighted user reads one off the screen — no special handling, no image of a QR code you cannot use.

Two things happen before you ever see the address. It is checked against the network's own checksum, and if it fails that check it is refused outright rather than shown with a warning — an address that is wrong by one character sends your funds nowhere recoverable, and that is not a risk to leave to the reader. And its capitalisation is preserved exactly, because on several networks the pattern of capitals is the checksum, so an address helpfully lower-cased in transit is a broken address.

Kraken is the first venue behind this. Note that the terminal reads deposit addresses; it does not create them. If you have never deposited that asset on Kraken before, no address exists yet and you will be told so — generate the first one once on kraken.com under Funding → Deposit, and the dialog will find it from then on.

Deposits need your own API key, so this is a desktop-app feature; the hosted browser terminal has no wallet.

Moving funds off a venue is not available. A withdrawal path exists in the code and is switched off. It is the one operation in this terminal where a mistake loses money directly rather than through a trade, and it will not ship until it has been run end to end against a live venue by a person. Until then no button renders, and the code refuses before any request could leave your machine.

Fills when you are not looking

A fill, a stop or a take-profit on the chart in front of you is spoken there. On any other chart, or with the browser closed, the same sentence arrives as a system notification, and the tray's 30-minute silence never covers it. The chapter Monitoring When You Are Not Looking has the rule, the two monitoring switches, and the tray icon.

OCO pairs: one cancels the other

The dashboard offers an OCO pair — two resting orders where whichever executes first cancels the other, announced like any cancel — whenever the venue the chart is reading can pair orders natively (Binance today) or you are in paper mode on such a venue. A sell pair brackets an exit: a take-profit limit above and a protective stop below; a buy pair brackets a breakout: a stop above and a pullback limit below. The form is one side, one quantity, and the two prices, and it refuses an inverted layout, or two legs at the same price, out loud rather than resting a pair that would fill instantly. On a live venue the pairing is held by the exchange ("OCO pair resting, linked by the exchange"); in paper the simulator enforces it, and the pairing survives a restart. If the second leg fails to place, the first is cancelled and the cancel is verified — you are never told a pair is gone while one leg still rests. The panel does not appear on a venue that cannot pair, in paper mode or otherwise, so paper is always a rehearsal for something the venue can actually do.

Reading the order book

Press Alt+B to open the order book for the current symbol; your screen reader announces it as "Order Book — {symbol}". The toolbar's Order book button appears only when the current provider has a book to show — the nine venues that stream depth (Binance, Bitstamp, Coinbase, Gemini, Kraken, Kraken Futures, MEXC, Alpaca, Polygon) and Interactive Brokers — so on a data feed with no book the button is simply absent, the way Deposit is absent on a broker with no wallet. Alt+B still opens the dialog anywhere, and on a provider with nothing it says so. It presents the resting buy and sell interest as two columns — "Bids (Buy Orders)" and "Asks (Sell Orders)" — each listing up to twenty price levels with their size and a running cumulative total, under a summary line giving the best bid, best ask, and the spread as both a number and a percent (for example "1.50 (0.004%)").

You read the book by Tab: every price level is focusable, and landing on one is announced as "Bid {price}, size {quantity}" or "Ask {price}, size {quantity}", so you can walk down the bids to feel where demand thins out or up the asks to find a wall of supply. If your provider streams the book it updates live as you read; otherwise a "Refresh" button pulls a fresh snapshot. It updates quietly rather than narrating every change, so it is a place you go to read depth deliberately.

A trade from start to finish

Putting it together, in paper mode: you are watching BTC/USDT on Binance and decide to buy a pullback. You press Alt+T, hear "Paper (simulated)", and Tab through the ticket — BUY, quantity 0.5, type Market, stop loss 42,180, and a Trailing take-profit of 1.5 percent with an activation at 43,000. You activate "Submit Buy Order"; because you are in paper it places at once, and you hear the earcon and "Order filled. Bought 0.5 BTC/USDT at 42,500." Price climbs through 43,000, arming the trailing take-profit, and then pulls back: "Trailing take profit hit. Sold 0.5 BTC/USDT at 43,260. Profit 380.00." You open the History tab and the trade is there with its price, profit, and fee. Had price instead fallen to 42,180 first, you would have heard "Stop loss hit. Sold 0.5 BTC/USDT at 42,180. Loss 160.00." — and either way, when you want out early, the Close button on the Positions tab flattens you with one press.

The History tab is real on every trading broker — Binance, Kraken, Tradier, Alpaca, Coinbase, and Schwab all report your recent fills with price, quantity, and fees. And the terminal watches your back across restarts: when it reconnects to a broker it compares your positions against what you held last session, and anything that closed while the app was off is announced — "While you were away on Kraken: BTC/USD position closed. Sold at 92,300. Profit 1,150." A stop that fires overnight is never a silent surprise. Cancelled orders speak too ("Order cancelled for {symbol}") — no order leaves the book without you hearing it.

Risk Management

The main objective is to preserve capital. A trader who keeps their account survives long enough for skill and luck to matter; a trader who loses it has no next trade. Everything in this chapter follows from that one sentence, and most of the terminal's trading features exist to serve it: paper trading, the stop-first quick trade, the risk/reward tool, the live review, the Journal. This chapter is not a course in trading. It is the handful of practices that keep one bad decision from becoming the last one, and where each of them lives in the terminal.

Decide what you can lose before you decide what you can make

Every trade has a price at which you were wrong. Name it before you enter. That price is your stop loss, and the distance from your entry to it, multiplied by your position size, is the amount you are risking. Naming the stop first turns "I think this goes up" into "I think this goes up, and if it trades below 42,100 I was wrong and I am out for 160 dollars".

The terminal enforces this in one place: the quick-trade keys refuse to place anything until you have set a stop with Ctrl+Alt+Shift+X, and the stop always travels with the entry. Nothing stops you typing a ticket in the dashboard with the stop field blank, but the protective-level validator will refuse a stop on the wrong side of price, because a stop above a long entry is not protection, it is an order to close the position on the next bar.

Size from risk, not from conviction

The standard rule, and the one every position-sizing calculator implements, is to risk a small fixed fraction of the account on each trade: half a percent to two percent. On a 100,000 account, one percent is 1,000. The position size follows from that budget and the stop distance:

quantity = risk budget ÷ (entry price − stop price)

Entry 43,700, stop 42,100, budget 1,000 gives 0.625 units. A wider stop means a smaller position for the same risk; a tighter stop means a larger one, and that second half is the trap, because a stop three ticks away sizes a position many times the account. The quick trade's Position value mode (the default) avoids the trap by making the percentage the size of the position instead; its What I lose if the stop is hit mode is the rule above. Choose in Settings → General → "Quick trade: the risk percentage means", and listen to the summary either way: it speaks the position value and the loss at the stop, and the one you did not choose is the one that surprises people.

The dashboard's Size button does the same arithmetic from the "Risk % of balance" field and the stop you typed, rounding down, from the cash in the destination account's quote currency and nothing else. A strategy's risk plan sizes the same way, half a percent by default.

Reward has to be worth the risk

A trade that risks 1 to make 1 needs to be right more than half the time to pay for fees and slippage. A trade that risks 1 to make 3 can be wrong twice for every time it is right and still come out ahead. That ratio is what the Risk/Reward tool (Alt+Shift+P) measures: place the entry, the stop, then the target, and it says "Risk to reward, 1 to 3.00". If the number is poor, the answer is usually not a nearer stop, which just makes you wrong more often, but a different trade.

Two of this project's own findings belong here, because they are easy to get backwards. Fixed targets and scale-outs tested badly in the Strategy Lab: returns are concentrated in a small number of large wins, and an exit rule that caps the right tail throws most of the return away. And pyramiding, adding to a winner on the way up, lost to simply taking the full position at entry on 47 instruments out of 47, because every add buys later at a worse price. If a bigger position is what pays, take it at entry where the price is best, and spend the risk budget on size rather than on a schedule.

Practise where the money is pretend

Paper trading exists so that the whole workflow, the keys, the announcements, the feeling of a stop firing, is familiar before it costs anything. The simulator fills at real live prices, charges a fee, refuses what you cannot afford, models shorts with collateral and a liquidation price, and runs brackets, one-cancels-other pairs and trailing exits. Everything in the Trading chapter behaves identically in paper and live. The one thing it cannot teach is how you behave when the loss is real, so treat the first live trades as paper trades with a real balance: small, stopped, and reviewed afterwards.

Know which account you are on

The most expensive mistake available in this terminal is placing a live order you thought was paper. The terminal names the environment in four places: the status bar badge (PAPER or LIVE), the spoken sentence when the dashboard opens ("LIVE account. Orders here are real money."), the "API key for this order" dropdown, and the live review before every real order. Any key not explicitly marked Paper is treated as live, and a key marked Paper on a venue that has no practice environment is refused rather than quietly routed to real money. Make confirming the environment the first thing you do in every trading session.

Leverage, correlation, and the warnings you will hear

Leverage multiplies both directions and adds a liquidation price: the level at which the exchange closes the position for you. The live review compares that level with your stop and warns if the exchange would act first, and the terminal keeps warning, once per approach, while an open leveraged position drifts toward liquidation. Correlated positions are one bet: Bitcoin and Ether, gold and silver, SPY and QQQ. The review reminds you when an order stacks onto an open position in the same sector, against a budget of two percent per sector. Neither warning blocks the order. The terminal informs; you decide.

Write it down, and read it back

The Journal (Ctrl+Alt+Shift+J) records every fill, every stop, every alert and every strategy setup with its rationale, and the History tab holds every fill with its price, profit or loss and fee. A losing streak reads differently on paper than it felt in the moment. The strategy library's Evidence column is the same habit applied to rules: a strategy that says Untested or Falsified is telling you what to expect.

What the research says about edges

This project has tested a great deal of conventional technical analysis in its Strategy Lab, and most of it has come back null: random horizontal lines were respected as often as real swing levels, Fibonacci ratios did nothing across 355,000 tests, and swing structure used as an entry trigger did not beat buy-and-hold once the entry was placed where it could actually have been known. The indicators and formations in this terminal are still worth hearing, because they are how you build a picture of the chart. Whether acting on them makes money is a separate question, and one to answer with a control before it is answered with capital. Use the tools to understand where you are; get your edge somewhere you have tested; and keep the stop.

Automation

Three features let the terminal watch and act so you do not have to stare at every bar: alerts that speak up when a condition you set is met, strategies that evaluate a rule set and propose (or place) trades, and custom scripts that let you bring your own indicator logic. The strategy tools are research instruments — treat them as exploratory, as the in-app banner says — but the alerts are an everyday convenience.

Alerts

Press Alt+J for the alerts manager. The top of the dialog is a short form for adding one: a Name (the placeholder suggests "e.g. Price crosses 50000"), a Target of Price, Candle, Indicator, or POC, a Condition — crosses above, crosses below, enters a zone, exits a zone, or changes direction — a Price Level to test against, and a Delivery choice of Speech, Earcon, or Both. Fill it in and activate "Add Alert"; the alert joins the "Active Alerts" list above, where each shows its name, target, condition, and level with a Delete button. There is no separate edit step — to change an alert, delete it and add a new one. At the bottom of the dialog, beside Close, is Delivery settings — the email, Telegram, webhook and browser- notification channels that carry alerts to you when you are away from the terminal. It opens as a second view of the same dialog, so Escape still closes one thing; "Back to alerts" returns you to the list, and to the button you left from.

The same panel holds Events you cannot see — one switch, on by default, that governs every system notification the terminal sends, with the timeframe floor beneath it. The Monitoring chapter explains the rule those follow. The hosted site has no desktop to reach and shows its browser-notification (Web Push) controls instead.

When an alert fires it reaches you immediately. Per its Delivery setting it speaks, interrupting whatever is being said — "{name}: crossed above {level}. Current value {value}." — and/or plays an alert earcon, and the event is written to the Journal so you can read it back. Alerts are never gated by your speech or sonification toggles — a condition you asked to be told about will always tell you. If you have set up email, Telegram, or webhooks under Delivery settings in this same dialog, fired alerts are sent there too, so you can be notified away from the keyboard. Webhooks are a named list: add as many as you like — "BTC channel", "Gold channel" — each pointing at a Discord webhook URL, a Slack incoming webhook, or any custom HTTPS endpoint, and each alert chooses which one it posts to from a dropdown when you create it. That is how a Bitcoin alert lands in your #btc Discord channel while a gold alert lands in #gold. Alerts are also scoped to a symbol now: a new alert defaults to the chart it was created on and only evaluates there, so a BTC alert no longer fires against whatever chart happens to be on screen (choose "any symbol" if you do want that).

On the hosted terminal, alerts run only while your browser has the terminal open. The server-side evaluation described here in earlier versions of this manual, and the browser notifications (Web Push) panel that went with it, are switched off on the hosted site — deliberately: it is a paper-only terminal, and an alert that arrives after you have closed the browser is not something you can act on there. The code for both is present and can be turned back on for a future hosted tier; today neither renders and neither runs. Email, Telegram and webhook delivery all still work, from your open browser, as does everything else in the Delivery settings panel.

If you want alerts that keep working with the browser closed, that is the local web host on your own machine — see the chapter Monitoring When You Are Not Looking. While a browser is connected, the in-session pipeline owns the alerts it can see and the background monitor takes the rest, so an alert on a symbol with no tab open is watched either way; the monitor reports and never acts.

For conditions a single rule can't express, switch on Advanced condition in the add-alert form. The same rule-tree builder the strategy composer uses appears: groups (AND, OR, NOT, or a Score threshold) over leaves that test indicator components, with optional higher timeframes per leaf — "RSI below 30 AND price above the 200-day EMA" is three clicks. The alert fires the moment the whole tree first becomes true, then re-arms when it goes false; turn on repeat-while-active and it re-announces on your cooldown while conditions hold. Score trees speak their score ("conditions met, score 7 of 9"). Delivery channels and symbol scoping apply unchanged. One requirement: leaves reference indicators by code, so add the indicator to the chart before building conditions on it — which is also why an advanced alert evaluates only while its chart is open (in a foreground or background tab), never server-side with the browser closed. The terminal says so when you create one.

Strategies

Press Alt+S for the Strategy Manager — labelled EXPERIMENTAL, with the standing caveat that backtested results do not guarantee live performance. It opens on a row of tabs: Library (your saved strategies, each with Start/Stop), Build Setup (the composer), Active (running strategies, with Pause/Resume), Backtest, and Custom Script.

Your library starts empty, and that is deliberate. This application ships the tools — the chart, the indicators, the condition builder, the backtester — and leaves the choice of what to trade to you. It does not pick a strategy for you and it does not arrive pre-loaded with someone else's. Earlier versions seeded thirty strategies from the project's own research on first launch; that was removed, because a shelf of strategies the application put there itself reads as advice, and of those thirty only one had ever been tested against a proper control while six had been tested and failed. If you upgraded, nothing was taken away — your library still holds whatever it held before.

Two ways to fill it. Build one in Build Setup, or import a strategy file.

Importing. At the bottom of the Library tab, Import strategies takes a .json strategy file — either choose the file or paste its contents, whichever suits you — and adds what it contains. The rules are fixed and worth knowing: importing never overwrites a strategy you already have (a file that contains one you already hold is skipped, and you are told), it never starts anything (everything arrives stopped, however the file was saved), and it refuses strategies that carry program code, which belong in the Custom Script tab where you paste the code yourself. Afterwards you hear the whole outcome in one sentence — how many were imported, how many skipped, how many rejected, and how many are set to place orders rather than only suggest them.

The Evidence column. Every row in the library says how far that strategy has actually been tested: Untested, In-sample only, Walk-forward, Control-tested, Fragile, or Falsified — with the detail alongside it, in the Description column: what it was tested on, which controls were run, and the verdict in a sentence, negative verdicts included. Strategies you built yourself read Not recorded, which is simply the truth about them. The point of the column is that a tested strategy and an untested one should never look alike in a list.

To build one, the Build Setup tab gives the strategy a Name and a Side (Long or Short), then a condition tree you assemble from "+ Group" (AND/OR/NOT) and "+ Leaf" buttons — each leaf picks an indicator, an operator, and the component to test, with an optional timeframe for multi-timeframe rules. Beneath it a risk plan sets the stop source (a percent of price, an ATR multiple, below a swing low, or a fixed price), a take-profit ladder of one or more rungs, and a stop buffer. Adding the finished setup to the engine marks it to re-load on the next launch, so your strategies survive a restart.

A running strategy talks to you as its state changes. When its conditions line up it rings a setup bell — a bright ascending chord for longs, a heavy descending chord for shorts, unlike any other sound in the terminal — and speaks the complete trade plan: "Long setup, {strategy name}, score 0.85. Entry 50,000, stop 49,500, target 1 51,000, target 2 52,500 (R:R 2.50). {why}." Every ladder rung is spoken, so you can hand the order to the strategy or place it yourself, manually, from what you just heard. — and if the entry is conditional you will hear it arm ("waiting for {trigger}") and then report when the trigger is reached. While the setup holds it heartbeats a quieter reconfirmation, and if a condition drops away it names what fell off. It also tells you when it is not yet ready: "indicators warming up — {n} of {m} bars loaded. Signals begin once warm." Every one of these lands in the Journal's Setups filter for review.

The Lab tab brings the research workflow in-app. Walk-forward windows slices the loaded data into equal chronological windows and backtests one strategy in each — hear whether an edge holds across regimes or lived in one lucky stretch. Compare all strategies backtests every saved strategy on the first and second half of the data and ranks them; SURVIVOR means the 95% confidence lower bound on per-trade reward-to-risk is positive in both halves with at least five trades each — the same statistical gate the offline research harness applies, because a single positive average on a dozen trades is indistinguishable from luck. Rankings favour the weaker half: a strategy is only as good as its weaker regime. Results are tables your screen reader can walk, plus a spoken verdict.

The Backtest tab runs a strategy over history with realistic settings — starting capital, commission, slippage, and a warm-up period you can auto-detect — over an optional date range, including one-press "first half / last half" buttons for walk-forward testing. Run it and the results read back as trade count, win rate, total P&L, max drawdown, Sharpe ratio, and an expandable trade log of each entry, exit, and reason. Because the whole feature is experimental, read those numbers as a study of the rules, not a promise.

Background monitoring

Alerts and strategies on tabs you are not looking at, and on charts with no browser open at all, are covered in the next chapter, Monitoring When You Are Not Looking. The short version: two switches in Settings → General, "Keep watching other tabs" and "Keep monitoring when the browser is closed", both off by default; background strategies announce and never place orders; Ctrl+Alt+Shift+M reports the status of both.

Custom scripts

When the built-in indicator set doesn't have the one you want, you can write your own. Press Alt+Comma for the Custom Scripts panel. If you can write a little C# — or you have a PineScript indicator from elsewhere — you can add an indicator that behaves, and sounds, like any of the built-ins. This section walks through writing one from scratch; deeper authoring (multi-output indicators, full control over how each output is voiced, or packaging a compiled plugin) is covered in the SDK guide and docs/PLUGIN_AUTHORING.md.

What a custom indicator is. Under the hood every indicator is a small class that takes the price history and returns one or more arrays of numbers — one number per bar, per line it draws. Your script implements a contract called ICustomIndicator, which is just six things: an Id (a short stable code), a DisplayName (what you'll hear it called in the indicator list), the ComponentNames (one name per output line), the DisplayTypes (how each line is drawn — and, importantly, how it is heard), a set of DefaultParameters, and a Calculate method that does the maths. The panel pre-fills a commented skeleton of exactly these members when you start a new script, so you are never staring at a blank page.

Two of those deserve a word, because they are where accessibility lives. The DisplayTypes you choose are not only about drawing — they decide the sound. Declare a line as an Oscillator and the terminal voices it around its zero line, so you hear it cross from negative to positive without reading a number; declare it as a Dot or Arrow and it becomes a sparse marker you can jump between with Ctrl+Left and Ctrl+Right. And the DefaultParameters you expose show up in the indicator's Properties dialog (P), so you — or anyone you share the script with — can retune the period or a threshold later without editing code.

A worked example. Here is a complete, working custom indicator: a Rate-of-Change oscillator that measures how far price has moved, as a percentage, over the last n bars.

public class RateOfChange : ICustomIndicator
{
    public string Id => "ROC_CUSTOM";
    public string DisplayName => "Rate of Change";

    // One output line, drawn — and voiced — as an oscillator around zero.
    public string[] ComponentNames => new[] { "ROC" };
    public ComponentDisplayType[] DisplayTypes => new[] { ComponentDisplayType.Oscillator };

    // Appears in the Properties dialog (P), so the period is tunable without code.
    public Dictionary<string, double> DefaultParameters => new() { ["Period"] = 14 };

    public double[][] Calculate(ReadOnlySpan<Ohlcv> data, Dictionary<string, double> p)
    {
        int period = (int)p["Period"];
        var roc = new double[data.Length];
        for (int i = 0; i < data.Length; i++)
        {
            // No value until there are `period` earlier bars to compare against —
            // NaN tells the chart "nothing here yet" so warm-up bars stay silent.
            if (i < period) { roc[i] = double.NaN; continue; }
            double prior = data[i - period].Close;
            roc[i] = prior == 0 ? double.NaN : (data[i].Close - prior) / prior * 100.0;
        }
        return new[] { roc };   // one array per component name, each the length of `data`
    }
}

A few rules the example shows: Calculate receives the full history, oldest bar first; it must return one array per name in ComponentNames, and each array must be the same length as the input; and any bar you can't compute yet — the warm-up at the start — should be double.NaN, which the terminal treats as "no value" and skips in both the drawing and the sound. An indicator with two lines (say a value and a signal) just returns two arrays and lists two ComponentNames and two DisplayTypes.

The workflow. Press "+ New", give the script a name, and write or paste your class into the editor. "Save" keeps it in your library. Activating Compile builds it in the sandbox and tells you the result — "Compiled successfully. Indicator: Rate of Change" when it builds, or a read-aloud list of the compiler errors when it doesn't, so you can fix them and try again. Once it compiles, "+ Add to Chart" drops it onto the chart as its own pane, and from that moment it is an indicator like any other: Page Down to it, Up and Down through its components, set its waveform and bell in Properties (P), hear it in playback, and jump its signals with Ctrl+Left and Ctrl+Right. "Export .atpkg" packages the script to a single file you can share or back up.

Bringing in someone else's script. The Import controls take an .atpkg file from disk or pasted JSON. Because that is code from outside, the terminal treats it as untrusted and asks first — "Import untrusted script '{name}'? … It will be sandboxed, but review the code before pressing Compile." — so you always get a chance to read what you're about to run. It is imported but not compiled until you choose to; read the source first.

From PineScript. If you have a PineScript v5 indicator, paste it into the "Transpile from Pine Script v5" box and press Transpile. The terminal converts it to C# you can then review and Compile, reporting "Transpiled to C# — review and compile." and listing any conversion warnings. It handles a practical subset of Pine rather than the whole language, so always read the generated C# and check the warnings before relying on it.

The sandbox, and one platform limit. Compiling and running a custom script never endangers the rest of the terminal: the code runs in an isolated worker process with no file, network, or reflection access, limited to the charting and indicator libraries, and on Windows, macOS, Android, and Linux that worker is additionally locked down by the operating system itself. The one exception is iOS, which provides no such sandbox, so custom-script compilation is disabled there — the editor still works as a text editor, it just can't run; use the Windows, macOS, or Linux build for scripting. The full design is in docs/SANDBOX_DESIGN.md.

The sandbox is required, not optional. If the operating-system sandbox is missing — on Linux that means the bubblewrap package (the bwrap command) is not installed — the terminal refuses to run scripts rather than quietly running them unprotected, and the compile attempt reports exactly what is missing and how to fix it. On Linux, install your distribution's bubblewrap package and restart. If you genuinely want to run scripts without the OS sandbox (for example on a locked-down machine you fully trust), set the environment variable ACCESSIBLETRADER_ALLOW_UNSANDBOXED_SCRIPTS=1; the terminal will honour it, and it records a security event every time a script runs that way so the decision is never invisible.

The Strategy Lab

Everything above runs inside the terminal. The Strategy Lab is a separate, optional command-line research harness that ships in the same repository — the place where strategy ideas are tested before they earn a spot in the application. The division of labor is deliberate: the lab is for research, the terminal is for trading. Nothing you run in the lab touches your workspaces, keys, or positions, and nothing in the terminal depends on the lab being present.

The lab also owns the strategies. Its catalogue holds every spec this project has built — thirty of them, with their bracketed research tags ([v13], [v23] and so on) — each carrying a record of how far it was actually tested and what the verdict was. None of them is installed with the application. To put one in your own library you export it from the lab and import the file, which is the only route in:

dotnet run --project AccessibleTrader.StrategyLab -- catalogue list --verbose
dotnet run --project AccessibleTrader.StrategyLab -- catalogue export \
    --out my-strategies.json --id builtin.long.trend-baseline

catalogue list --verbose prints each strategy with what it was tested on, which controls were run, and the verdict — read it before you import anything. --status narrows the list to one evidence level (--status Falsified is a short and instructive list). Exporting in bulk by --min-evidence deliberately refuses to include anything recorded as Fragile or Falsified; if you want one of those, for instance to re-test it yourself, you have to name it with --id.

You do not need the lab to validate a strategy day-to-day: the Strategy Manager's Backtest tab has walk-forward first-half / last-half buttons built in, which is the same honesty check in point-and-speak form. Reach for the lab when you want the heavier machinery — testing dozens of gate combinations at once, or stress-testing across many rolling windows instead of one split.

The lab is run from a terminal window with the .NET SDK installed, from the repository root. dotnet run --project AccessibleTrader.StrategyLab -- help lists every command; the ones that matter:

• snapshot — download bar history for a symbol into strategy-lab-data/ (the lab always works from these saved snapshots, so experiments are reproducible).
• catalogue list / catalogue export — see what the lab holds, what the evidence for each spec is, and write a strategy file the terminal can import.
• cftc-cot — download CFTC positioning history for named contracts.
• run / walk — backtest one built-in strategy by its id against a snapshot; walk splits the data in half and reports each half separately, which is the minimum standard of evidence used throughout this project.
• battery — run the full grid of entry/gate combinations against a snapshot and report which cells survive both halves with statistical confidence.
• rolling-window — the strictest test: the same battery across many overlapping windows, reporting how consistently each cell wins, not just whether it won once.

A typical session: fetch a snapshot, walk a strategy id you are curious about, and read the two halves side by side. If the first half and the second half disagree, the lab just saved you from a strategy that only worked in one era — which is precisely its job. All output is plain console text, fully readable with a screen reader, and per-trade CSVs are written for deeper review.

Monitoring When You Are Not Looking

Normally the terminal watches the chart in front of you. Two settings extend that: one to the other tabs you have open, one to the time the browser is closed altogether. They are different switches for different situations, and this chapter covers both, plus the rule that decides how each announcement reaches you and the tray icon that drives the browser-closed half.

Where each announcement goes

The terminal decides how to tell you something by asking one question: is this about the chart in front of you?

• If it is, it is spoken where it always was: through your screen reader, with its earcons. It is never also a system notification, because you are already being told. Minimising the browser does not change this; the page is still loaded and your screen reader still reads it.
• If it is not, it arrives as a system notification: a bar closing on another tab, an alert or a fill on a market with no tab open, and, once the browser is closed, every event there is. A notification is the only channel that can reach you when the words have nowhere else to go. It goes through your desktop's notification daemon on Linux (MATE, GNOME and KDE all show it and Orca reads it), the Action Center on Windows, Notification Center on macOS. On a machine with no notification tool at all, or when a notification is sent and does not arrive, the same sentence is spoken aloud instead.

One switch governs the second bullet: Alt+J → Delivery settings → Events you cannot see → "Send me a notification". It is on by default. Beneath it sit Shortest timeframe to announce, a floor so a one-minute chart does not become a notification a minute (it covers bar closes and the narration ladder, never alerts or fills), and Also speak bar closes from other tabs, off by default, which adds speech to the earcon a background tab's bar close already plays, waiting for the current sentence to finish.

Three things ignore all of this. Errors and order outcomes come through every mute and every snooze. The soundscape never mixes across tabs: playback, navigation tones and the sonification bed belong to the focused chart only.

Keep watching other tabs

By default only the chart on screen is live; switch from a Bitcoin tab to a gold tab and the Bitcoin alerts and strategies go quiet until you switch back. Settings (F12) → General → "Keep watching other tabs" lifts that. Every other open tab is then re-fetched on a polling cadence (every 30 seconds by default, adjustable, floor 10), its indicators recomputed, and its symbol-scoped alerts and running strategies evaluated on the fresh bars. "Live-stream other tabs" keeps up to eight background tabs on real streaming data instead of the poll where the exchange allows it. It is off by default because it spends your provider's request budget, and it is a desktop and local web host feature; the hosted site stays single-chart.

What you hear follows the rule above: a background tab's alerts and setups reach you at full priority, as notifications prefixed with the symbol ("BTC/USD: crossed above 50,000"), and its email, Telegram and webhook deliveries fire as if the tab were on screen. Two rules keep it honest. An alert or strategy is evaluated by exactly one side at a time, the live pipeline while its tab is focused and the background monitor the moment you switch away, so nothing fires twice. And background strategy signals are announce-only: a strategy in Auto mode speaks its setup from a background tab but never places an order there. Its managed exits still run, because a stop that only works while its tab is on screen is worse than no stop.

Order placement is not the same as order watching. A chart you have a position or a resting order on is watched whether or not this switch is on, across tabs, after the tab is closed, and after a restart: a stop fires and a fill announces wherever you are.

Keep monitoring when the browser is closed

On your own machine, the local web host is a server that outlives the browser tab. Settings → General → "Keep monitoring when the browser is closed" puts that to work. It is off by default.

When you close the last tab, the terminal says so. A notification arrives a few seconds later: "The browser is closed. The terminal keeps running: alerts, order fills, bar closes and narration for your saved charts arrive here as notifications until a browser connects again." If the switch is off, the same notification says that nothing is being watched and names the switch. A reload does not trigger it, and three tabs closed together produce one.

With it on and no browser connected, the terminal checks about once a minute and tells you about:

• Alerts. Every alert that names a symbol and provider keeps evaluating; the watch list is your alert list. Indicator, point-of-control, trend, zone and condition-tree alerts are evaluated too, on a real chart the monitor rebuilds from your last saved session. The two it cannot watch are an alert scoped to "the current chart" (no symbol to fetch) and a POC alert on a chart with no profile; the terminal says so when you create one.
• Order fills, stops, take-profits, rejections, cancels and replacements on every venue you hold a key for with open work. Reported, never acted on: the monitor places, moves and cancels nothing.
• Bar closes on your saved tabs (up to eight), above the timeframe floor.
• The narration ladder, the same reading a bar close gets in the browser, for every series you flagged with N, in the same sentence as the bar close and led by the symbol.

Bar closes and the ladder follow your saved tabs; alerts follow your alert list. Close a tab and its bar closes stop being watched even if an alert on that symbol keeps firing. Strategies are not evaluated with the browser closed at all.

If the monitor's own polling fails (a provider down, a bad key) it says so once, and again when it recovers. The alert sound is replaceable: drop your own WAV at sounds/alert.wav in the app-data folder. Pair the terminal with a systemd user service and it listens from login to shutdown.

Ctrl+Alt+Shift+M speaks where things stand: which of the two switches is on, which tabs are watched and how fresh their data is, how many alerts and strategies are armed, and whether anything will be watched once the browser is closed.

The system tray icon

On a local machine the terminal puts an icon in the system tray (the notification area). Its accessible name carries the unread alert count ("Accessible Trade Terminal — 3 new alerts") and its menu, which your screen reader navigates like any menu, has seven items:

• Restore workspaces to browser reopens the terminal in your browser; the session resumes.
• Show recent alerts speaks how many are waiting and opens a plain page where each alert has Mark as read and Dismiss, plus Mark all read. Bar closes and narration that happened while you were away are not filed here; the alerts are.
• Silence alerts and bar closes for 30 minutes quietens alerts, bar closes, the ladder and the monitor's own reports. Order fills, stops and take-profits always come through. Alerts keep being evaluated and still reach your email, Telegram and webhooks. The item then reads "Resume alerts and bar closes" with the minutes remaining.
• Connection status speaks whether monitoring is on, how many alerts are armed and unread.
• Copy terminal address copies the local URL for another browser or device.
• Turn background monitoring on / off is the "Keep monitoring when the browser is closed" switch, reachable without a browser. It is not the "Keep watching other tabs" switch.
• Exit terminal shuts the server down.

On Linux the tray uses the freedesktop StatusNotifier protocol and its menu is exposed to your screen reader; on Windows it is the standard notification-area icon; on macOS the web host provides the actions but no icon. If your desktop cannot host a tray icon the monitoring above still works.

The native apps

The Windows app has no browser to close, so its equivalent is the window. Closing it with the X or Alt+F4 minimises to the notification area by default and a notification confirms the terminal is still running; from then on every event, including the focused chart's bar close, arrives as a notification because nothing is being read aloud on screen. Restore or Quit from the tray icon. Settings → General → Minimize to tray on exit turns this off. This path is built and tested in code but has not yet been run on a Windows machine.

The native apps have no headless monitor: with the window open they watch the focused chart and, with "Keep watching other tabs" on, the other open tabs, and nothing else. An alert on a symbol with no tab open is watched by nobody on the Mac and mobile apps, and the Mac app has no tray and no minimise-to-tray. For browser-closed monitoring, run the local web host.

The hosted website runs none of this: it is a paper-only terminal, its alerts evaluate only while your browser has it open, and it never shows a tray.

Customizing

The terminal is meant to be shaped to how you work and what you can hear. This chapter covers the three places you do that shaping: the settings dialog for global preferences, the sound designer for the audio palette, and tabs and workspaces for arranging and saving your charts.

Settings

Press F12 for the settings dialog. It has nine tabs on the desktop apps — General, Speech, Narration, Sonification, Braille, Appearance, Keyboard, License, About — and eight in a browser, where the Braille tab is absent because a Dot Pad connects to the machine running the terminal, never to the viewer's browser. Left and Right arrows move between them, and a Search settings box at the top jumps to any setting by name.

Save is what saves, and Escape is Cancel. Every control in this dialog holds your change until you press Save; Save writes them all at once and says "Settings saved", and Cancel — or Escape, or a click on the backdrop — drops them all and says "Settings discarded". You can change your mind about anything in here without consequence, and you never have to wonder whether a box you ticked and then thought better of has already taken effect.

There is one deliberate exception, on the Appearance tab. The theme, the text size and the visual accommodations (chart formations drawn, visual earcons, colour-vision-safe colours, hollow up-candles) apply the moment you pick them, because a visual setting you cannot see while you decide is a setting you cannot judge. Cancel restores them as well, and says "Settings discarded. Appearance restored" so you know it did.

Speech and Narration are split by a rule worth learning, because it tells you which tab to go to without hunting: Speech is how the terminal says what you asked for, and Narration is what it says when you pressed nothing. So Speech holds whether it speaks timestamps and when, whether it reads column headers, the order it reads a value and its name in, and which voice talks on this device — every one of them a change to the answer you get back from a key you pressed. Narration holds the three things that speak on their own: the new-bar announcement, signal narration on a bar close, and narration during playback.

Two settings you might expect on Speech or Narration are on General, under Analysis: Describe candle patterns and Describe chart patterns. Neither belongs to a single trigger — each changes what the arrow keys say and what a bar close says and what playback says — so filing either by trigger would put it on a tab whose rule it breaks. (That sentence was aspirational for the candle switch until 2026-09-04, when the arrow keys started naming candle patterns at all; it is now simply true.) They are filed by what they are, an analysis the terminal either performs or does not, and they sit next to each other because that adjacency is what makes the difference between them legible: a candle pattern is one to three bars (engulfing, harami, doji, hammer), a chart formation is tens of them (double top, head and shoulders, triangle). Both answer to "narration" and "speech" in the settings search box.

Sonification mirrors F3 and holds the sound theme, the sound-under-the-mouse option and the two earcon families described below. General is where you switch paper trading on and choose what the quick-trade risk percentage means, plus the session resume, the two background-monitoring switches, the tray behaviour on Windows, the touch toolbar, analysis (Market Structure on new charts and the two pattern switches), magnet snap, the panning step, a CSV export — and the factory reset, below. Braille enables the tactile display. Appearance is the theme — most relevant to a sighted collaborator looking over your shoulder — and the visual accommodations. What can be shared, and how, is under Sharing your setup below. Keyboard rebinds any command, and those rebinds save the moment you make them, the one exception to the Save rule.

Save and Cancel. The dialog commits when you press Save, and only then. Escape, the Cancel button and a click outside the dialog all discard — that is the rule everywhere in the terminal now, and it was not always true here: until 2.6.0 this dialog had a single button reading Close, and closing is what saved. A few controls still take effect the moment you change them, and each of them says so as you do: the theme picker, text size and the visual accommodations (previewed, and put back on Cancel), the paper-account reset and the factory reset (commands, not edits), and keyboard rebinds.

The two earcon families (Sonification tab)

Earcons — the short tones that stand in for a sentence — come in two families, and each has its own checkbox under Shift+F3, which remains the single mute over both.

• Market earcons are the ones about the market: an alert firing, a new bar opening, a strategy setup arming or reaching its entry.
• Interface earcons are the ones about the terminal: the edge of the chart or of a series, a mode toggled with F2 or F3, an action that succeeded or is being retried, the connection changing state.

Both start on, so an install that never opens this tab sounds exactly as it always did. They are worth separating because they fire at wildly different rates for wildly different reasons: the boundary tone sounds on every further arrow press at the edge of a chart, and a setup bell might fire twice in a session. One switch used to govern both, so quietening the first meant losing the second. Every interface earcon is also spoken, so turning that family off costs you the beep and nothing else.

Neither switch reaches error tones or order outcomes — a fill, a stop, a take-profit, and every error. Those sound with both families off and with Shift+F3 muted. There is no compensating channel for an error you cannot see, and money moving is neither a market observation nor an interface confirmation.

Factory reset (General tab)

"Reset all settings to defaults" arms a confirmation the same way the paper-account reset does, and for the same WCAG 3.3.4 reason. It puts back to shipped defaults: every setting in this dialog, every keyboard rebinding, your own themes, your sound patches and earcon assignments, and the colours and sounds you gave individual indicators.

It keeps your API keys, your paper trading account and its history, and your saved workspaces, and the confirmation says so — because "all personalization will be lost" is a sentence most people will read as "including my broker logins", and being wrong about that in the frightening direction is what stops someone using a button they need. Your keys are credentials rather than preferences and nothing on this machine could rebuild them; the paper account is a trading record with its own reset a few rows above; your workspaces are documents you named and can delete by name.

Every part of the reset is attempted even if one of them fails, and you are told how many failed rather than simply "done" — a reset that stopped at the first problem would leave you with a keyboard from one era and preferences from another and no way to tell which. Restart the terminal afterwards so every part of it re-reads the defaults.

The Narration tab

"What the terminal says on its own, without you pressing anything." Six switches. Most are on to begin with, and that is not the terminal being chatty at you: none of them lets anything new through on its own. Signal narration only ever speaks about series and components you flagged yourself with N, so on a chart where you have flagged nothing it has nothing to say; and playback narration is on because it is what the terminal already did.

• Announce new bars — the rolling "Close … New bar …" heartbeat of the live candle. It also gates the two forming-bar switches below.
• Narrate candle patterns as they form (on) — "hammer forming", spoken as the live bar changes shape, debounced to one announcement per change. Needs "Describe candle patterns" on General, and N on the candles.
• Narrate chart formations as they form (off) — a double top or triangle while it is still building, with the level that would confirm it, once per formation. Off because it is a new occasion to speak.
• Narrate signals on bar close — the master switch over auto-narration. N chooses which series and components speak; this decides whether any of them do. Turn it off for an hour of quiet without having to un-flag six indicators and then remember which six they were.
• Narrate during playback — the master switch for everything playback says beyond its own start, pause, speed and finish confirmations: the date or hour as the tones cross each boundary, the signals your narrated series print, and any chart formation that resolves. Off leaves playback as pure tones. Nobody had that switch before 2.6.0.
• Speak time landmarks during playback — the date or hour alone, on by default and subordinate to the switch above. It is the only thing that says where in time the tones are, and it is also a calendar read to you every few seconds; turning it off keeps the signals and the formation outcomes and stops the dates.

There used to be an eighth tab, Alerts, holding the email, Telegram and named-webhook delivery details. Those moved into the alerts dialog itself (Alt+J, then Delivery settings), next to the alerts they deliver — the webhook names you define there are the choices in an alert's Webhook list, and having the two a dialog apart meant closing the alert you were writing to go and define its destination.

Appearance and the theme editor

The appearance section is easy to skip if you never look at the screen, and worth a minute if you sometimes do, or if anyone ever looks at your screen with you.

A theme covers the whole window, chart and chrome together, so switching one never leaves a themed chart sitting inside a fixed grey frame. Classic is the default. Blackout is pure black with white text, for OLED panels and for anyone who finds any lit background tiring; High-contrast dark and High-contrast light are the two built for maximum separation. Paper is a real light theme — warm off-white, near-black ink — meant for daylight, projectors and printing. Amber CRT, Walnut, Midnight Blue, Classic, Soft Dark, Solarized and Braille fill in the rest. The high-contrast pair are deliberately not the default: they are accessibility tools, and greeting every new user with black-and-white made a finished application look like a debug harness. They are one setting away and unchanged.

Beyond picking one, three buttons open the theme editor and mean what they say. New theme starts a raw theme from a plain scheme — black chart, green rising candles, red falling candles. Clone theme copies whichever theme the picker shows, built-in or your own, into a new theme you can rename. Edit theme changes one of your own themes in place; on a built-in it stays where it is and tells you why ("Built-in themes can't be edited. Use Clone to make your own copy.") rather than vanishing. Whichever way in, you get all 33 colours the application can draw, grouped into sections you can jump between with your screen reader — top bar, chart area, candles, overlays, bottom bar, dialogs, text. The chart background and its gradient are the Chart top and Chart bottom colours there, and a Blend into one gradient group at the top of the editor colours the toolbars, chart and footer as one fade from a top colour to a bottom colour, six fields at once, each of which you can still change or revert afterwards. Every picker carries a description of what the colour actually affects, because "gridlines, minor" means nothing on its own, and every change is announced as you make it — "Rising candle set to #26A69A." — so you know it took without needing to see it. Where two colours you have chosen are too close to read against each other, the editor says so inline and leaves them alone; it never quietly corrects your choice. Reset all returns to the base theme, and Save and use keeps your version as a named theme of your own, which saves and loads as text you can send to someone else.

Before 2.6.0 the Appearance tab also held a chart background colour, a gradient, a window gradient switch and a bullish/bearish colour pair that were layered over every theme. Those are gone: a colour belongs to a theme now, and anything you had set there is ignored. To get the same look back, Clone the theme you use and set the colours in the editor.

The Analysis group on the general tab holds one preference worth knowing about: Add Market Structure (swing highs and lows) to new charts, which is off by default — a new chart is candles, volume and price, and nothing you did not put there. Tick it if you want swing labelling on every chart you load. It changes what happens on the next load and deliberately leaves charts you already have open alone — silently adding or stripping an indicator on a chart you were reading would be a worse surprise than the setting waiting one load to take effect.

Visual accessibility options. Under the theme, the Appearance tab carries a group of visual accommodations. All of them are off by default — the terminal presents itself audio-first, and these exist for users who want a visual channel too. Each applies and saves the moment you toggle it:

• Visual earcons mirrors every sound cue — order fills, stops, take-profits, setups, errors, new bars, connection changes — as a brief on-screen badge naming the event, for deaf and hard-of-hearing traders or anyone working with the sound down. Each event fades in and out once; nothing ever flashes repeatedly.
• Color-vision-safe chart colors replaces the red/green up-down convention with blue for up and orange for down on candles and direction-coloured bars — the two hues remain distinct with deuteranopia or protanopia. While on, it deliberately overrides any per-indicator direction colours so one switch covers the whole chart.
• Hollow up-candles draws rising candles as outlines and falling candles filled, making direction readable by shape alone, with any colours.
• Toolbar captions puts a text label under every toolbar icon. Off by default because the captions cost the chart a third of its height on an ordinary window; the spoken name and the hover tooltip are the same either way.

The same group holds a Text size selector (85% to 175%) that scales the interface text throughout the terminal — browser zoom still works on top of it. And at the top of the whole dialog there is now a Search settings box: type a word like "speech", "theme", or "alerts" and matching settings are listed with the tab they live on; choose one and the dialog jumps there and focuses the control, so you never need to remember which of the tabs holds a setting.

Two further accommodations need no switch at all: if your operating system or browser is set to reduce motion, the terminal's animations and transitions are disabled automatically; and on touchscreens, buttons and tabs enlarge automatically to comfortable touch sizes. On high-resolution screens the chart image itself now renders at your display's native pixel density, so candles and axis text are sharp rather than softly upscaled. During playback, the highlighted bar on screen follows the audio bar for bar, so a sighted companion can watch what you are hearing.

The sound designer

Two layers of audio control sit underneath settings. The first is each indicator's properties dialog (P). Its Sonification tab has an Acoustics section for the component you pick there, and at the top of it a Sound Patch dropdown chooses the voice that component plays — any built-in patch (the bells and more) or any patch you have made yourself — with a ▶ Preview button beside it to hear it. The second and third dropdowns adapt to what the component actually is: price bars get Green (bullish) and Red (bearish) patches so rising and falling bars can sound different; zero-anchored histograms and areas get Positive and Negative patches split at their baseline; and bounded oscillators like the RSI get Above midline and Below midline patches split at the middle of their range (RSI 50). Plain lines show only the single patch. Whatever patch you choose, the overbought/oversold zone texture from the Reference Levels section still plays on top — a patch changes the instrument, never the zone cue. Leave a patch unset and the older manual controls — waveform, noise, volume — take over as a fallback, and "Save as Defaults" still makes your choices stick for the next indicator of that type.

The second layer is the sound designer, opened with Alt+W, where the patches themselves are built. It is now a general-purpose patch workbench rather than an earcon-only panel. A single patch can stack several oscillators, each with its own waveform (sine, square, sawtooth, triangle, or noise), level in the mix, frequency ratio (a harmonic multiple of the patch's base pitch — 2.0 is an octave up), and noise blend and noise colour (pink, white, or brown); the "Add Oscillator" button layers on another. A Mix section sets the base frequency, a frequency multiplier, and overall volume; an Envelope section chooses a sustained tone or a plucked Ping and its duration. The Preview button auditions the whole patch — noise and envelope included, not just the bare waveform as before. A patch you build here can be assigned to event earcons in this same panel, or to indicator components through the properties dialog above, and the link is live: edit a patch in the sound designer and every component and earcon using it updates at once. Patches saved by older versions still load unchanged. Think of the properties dialog as choosing which instrument each part plays and the sound designer as building the instruments.

New in this panel: Import WAV. A short single-cycle WAV file — the free public-domain AKWF collection is thousands of them, or one period of any recorded instrument — imports as a wavetable: a custom oscillator waveform that plays at any pitch and takes envelopes, noise, and layering exactly like the built-in shapes. A longer WAV imports as a one-shot sample for earcons and signal layers, played at its natural speed. Both appear in every oscillator's waveform list ("Wavetable: …" and "Sample: …") and survive restarts. If an import ever goes missing, patches that reference it fall back to a plain sine — audible, never silent.

The earcon list in this panel now includes the three reference-level cues — the crossing chirps (up and down separately), the approach ping, and the sustained-in-zone tone — so those can be re-skinned with any patch too. And a factory bank of ready-made instrument voices ships alongside the built-in bells: flute, clarinet, pipe-organ registrations, glass, and string-ensemble patches (they appear as "Voice: …" in every patch dropdown). They exist for the sound themes below but you can assign them anywhere by hand.

Sound themes. Settings (F12), Sonification, "Sound theme" assigns those factory voices automatically, one instrument per indicator family: with the Orchestra theme, price and moving-average lines are a flute, bounded oscillators like the RSI a clarinet, zero-cross indicators like the MACD a pipe organ, and band edges glass. The point is playback: press Space on a busy chart and you can tell which line is talking by its instrument alone. The Pipe organ and Strings themes voice the same families with different registrations of a single instrument, and Classic is the original pure-tone palette. A theme applies to indicators you add after choosing it, never touches candles, wicks, or volume (their sound carries size information that a fixed instrument would erase), and any per-component patch you pick in a Properties dialog always wins.

Tabs and workspaces

You can keep several charts open at once in tabs. Ctrl+T opens a new chart tab, Ctrl+W closes the current one, and Ctrl+Tab and Ctrl+Shift+Tab move to the next and previous tab — so you might hold BTC on the hourly in one tab and a stock index on the daily in another, and flip between them without reloading. A row of tabs sits just above the chart and is always visible, even when only one tab is open, so the "+" new-tab button is always there for the mouse. (On the Linux web host the browser claims Ctrl+T, Ctrl+W, and Ctrl+Tab for its own tabs, so there they are replaced: press Alt+Shift+N to open a new chart tab, and press Ctrl+Alt+Shift+T to move keyboard focus onto the tab switcher bar. Once the bar has focus, switch with the left/right arrow keys, Home and End, or the number row (1–9 jump straight to that tab); press Insert to open another tab and Delete to close the focused one. The bar is an ARIA tablist, so your screen reader announces each tab as you move. See the Platform Support chapter.)

Out of the box only the tab on screen is live — but if you want the others watched too, turn on background monitoring in Settings and their alerts and strategies keep evaluating while you work elsewhere, each announcement prefixed with its symbol. That feature has its own section in the Automation chapter.

A whole arrangement — every tab, its symbol and timeframe, its indicators and drawings — is a workspace you can save and restore. Ctrl+Alt+Shift+W saves the current workspace and Ctrl+Alt+W loads one back; because those are three-modifier chords the browser does not reserve them, so they work the same on every platform. Set up the charts and indicators you return to every session once, save them, and you are one shortcut from that whole layout the next time you sit down.

A workspace save captures every tab's identity (market, provider, symbol, timeframe), the indicator stacks with their settings and audio patches, your drawings, display toggles like Heikin-Ashi and log scale, pane heights, and the strategies that were RUNNING — each remembered with its symbol binding, its Suggestion/Auto mode, and whether it was paused, so loading the workspace brings them back exactly as they were. Alerts are not part of a workspace; they persist on their own the moment you create them. Bar data is refetched fresh on load.

You also do not need to remember to save at all: the terminal autosaves your session every thirty seconds and again when it closes, and the next start simply resumes it — you hear "Resumed your last session: N tabs" and you are back where you left off, tabs, drawings, strategies and all. This covers the browser-refresh case on the web build too, which previously lost everything unsaved. If you prefer a blank start, turn off Settings > Workspace > "Resume last session at startup"; explicit named workspaces are untouched either way.

Sharing your setup

Your settings are modular in one specific sense: each kind of thing lives in its own file in the app-data folder, so any of them can be copied to another machine or handed to another trader without the rest. The folder is ~/.local/share/AccessibleTrader/ on Linux and macOS and %LOCALAPPDATA%\AccessibleTrader\ on Windows. The pieces:

The two things people actually hand each other are the first two rows: a tuned soundscape and a theme. Until 2026-09-13 the Appearance and Sonification tabs also carried "Export Visual" and "Export Audio" buttons that promised a whole tab as one file and delivered only the current chart's per-component overrides; they are gone, and the patch and theme routes above are the ones that work.

A factory reset (General tab) clears everything in the table except your API keys, your paper account and its history, and your saved workspaces, and says so before it does.

The Tactile Display

Alongside speech and sound, Accessible Trader can drive a refreshable tactile graphics display — a pad of pins that rise and fall so you can read the shape of the chart with your fingers. All Dot Pad models are supported — including the Dot Pad X (the newest model) and the second generation — because they share the same graphics SDK, so the driver works across the family. Each is a grid of thirty by ten graphic cells — sixty by forty individual pins — with a separate twenty-cell braille text strip beneath it. (On-device testing so far has been on the second generation; the Dot Pad X uses the same SDK and is supported.) Tactile output does not replace the speech and the soundscape; it is a third layer you read at the same time, and it is the one that gives you the chart's form directly under your hand rather than as pitch over time.

Tactile support is a Windows feature today, because the device's graphics need the vendor's Windows driver; the Linux web host can't drive the pins yet (the vendor's Linux library has no graphics support — see the project docs for the current state).

Turning it on

Tactile output is off until you ask for it. Open Settings with F12 and, on the General tab, tick "Enable braille / tactile display output." With it on, the terminal looks for a connected Dot Pad as it starts, and keeps watching while it runs — so you can plug the display in at any time and hear it announce itself, "Dot Pad connected," and unplug it to hear "Dot Pad disconnected." Turning the setting back off stops all of this and skips device detection entirely. The toggle is deliberately opt-in: looking for the display means probing serial ports, which is the sort of thing you only want happening when you actually have one.

Reading the chart by touch

When a chart is loaded the pad shows it as two stacked panes — the focused series and the one above it in the cycle — with candles drawn as a body, a wick, and a gap so the bars are distinct under your fingers, and indicators drawn as the line, oscillator, or markers their type calls for. Beneath the graphic, the twenty-cell strip carries the live value of wherever your cursor is, switching for about a second and a half to the bar's timestamp each time you move with the left or right arrow before falling back to the value.

The graphic redraws when you navigate — change focus, switch panes, zoom, pan — but deliberately not on every live tick. Each tactile frame takes a second or two of physical pin movement, faster than ticks arrive, so redrawing on every tick would leave the pad permanently in motion and unreadable. The stable graphic stays put under your hand while the fast-moving live value rides on the strip, which is the right surface for it. Before any chart is loaded the pad reads "accessible trade terminal ready" and the strip "no chart loaded…".

The device keys

The Dot Pad's own four function keys and its panning keys are wired into the terminal so you can keep both hands on the pad. F1 speaks the focused series — "candles" for the price pane, or the indicator's name. F2 speaks the focused component within it. F3 speaks the chart's identity — symbol, timeframe, and provider. F4 freezes the graphic so you can study a frame without it redrawing under you, and frees it again on a second press; the strip keeps updating either way. The display's pan keys scroll the chart left and right exactly as the [ and ] keys do. Each of the function-key answers is also written to the braille strip, so you can read what F1 to F3 reported as well as hear it.

Setting up the hardware itself — installing the Dot Pad SDK, the supported connection, and calibration — is covered in the project's platform documentation rather than here.

Platform Support

Accessible Trader runs two ways: as a native desktop and mobile app on Windows, macOS, iOS, and Android, and as a self-hosted browser application — the Linux web host — that you reach in a browser such as Firefox. The keyboard navigation, the Hybrid Voice model, and everything in the preceding chapters are the same everywhere; what differs is the plumbing underneath and a single block of keyboard shortcuts on the web host.

• Windows — works with NVDA, JAWS, and Narrator; uses the WASAPI audio engine for the lowest latency; needs a full hardware keyboard for the shortcuts. The Windows build is also the one that drives a Dot Pad tactile display when one is connected.
• macOS — works with VoiceOver, uses the AVAudioEngine audio path, full keyboard support.
• Android — works with TalkBack and the AudioTrack engine; the keyboard shortcuts are available when a physical keyboard is connected.
• iOS — works with VoiceOver and AVAudioEngine; shortcuts require a connected hardware keyboard, and, as the Automation chapter noted, custom-script compilation is disabled because iOS provides no process sandbox.
• Linux (web host, in a browser) — works with Orca and other browser-compatible screen readers, with audio routed to your system through PipeWire or PulseAudio (or, on a remote/demo deployment, streamed to the browser).

Which version to use

The native app and the web host are not rivals so much as two doors into the same terminal, and which one you want is mostly decided by your operating system. On Windows and macOS, use the native app — it gives you the deepest integration: the lowest-latency native audio, a chart that redraws at full speed, and your credentials held in the operating system's own keychain. On Linux, use the web host — there is no native Linux build, and the web host is an excellent first-class client, speaking through Orca and playing through PipeWire or PulseAudio; it is also what powers the public chart demo on the website. And on a phone or tablet, the native app is the only option — there is no way to put the web host in your pocket.

Two capabilities tip the balance toward the native Windows app in particular. The tactile display described in the previous chapter is Windows-only, so a Dot Pad user wants that build. And the safety sandbox for custom scripts is fullest on the native desktop platforms. Everything else — every chart, indicator, drawing tool, alert, strategy, trade, and the whole Hybrid Voice model — works the same on both, so if you move between a Windows desktop and a Linux laptop you are using the same terminal in both places, with only the plumbing underneath changing.

The web host and the modifiers

The keyboard is the same on every head. Until 2026-09-05 it was not quite: Firefox and most browsers reserve several Ctrl+Shift+letter chords for themselves — reopen a closed tab, open a private window, and so on — and a web page cannot override them, so the browser host used to rewrite the desktop's Ctrl+Shift drawing chords to Alt+Shift at startup and the help had to explain the difference. The default profile now binds all the drawing tools and the detailed point summary to Alt+Shift+<letter> everywhere, so the difference is gone. A shortcuts profile saved before the change still carries the Ctrl+Shift chords; the desktop honours it as saved, and the browser host rewrites those legacy chords to Alt+Shift in memory. Chords with three modifiers — the AI Analyst, auto-narration, the Journal, save and load workspace — were never affected. A few single-Ctrl chords are reserved by the browser at an even deeper level — it acts on them before the page ever sees the keystroke, so they cannot be cancelled in-page at all. On the web host these are dropped from the bindings (so the Help dialog never lists a chord the browser eats) and the action moves elsewhere: open a new tab with Alt+Shift+N or the tab bar's + button instead of Ctrl+T; close a tab with the Close tab button beside it or by focusing the bar and pressing Delete instead of Ctrl+W; switch tabs by pressing Ctrl+Alt+Shift+T (a three-modifier chord browsers leave alone) to focus the tab switcher bar, then the arrow keys, Home/End, the number row, Insert (new) or Delete (close) instead of Ctrl+Tab. Pane navigation needs no such rule: it is Alt+PageUp / Alt+PageDown on every head, and Ctrl+PageUp / Ctrl+PageDown — which the browser uses to cycle its own tabs — is left unbound rather than reassigned. You never have to memorise which is which — the Help dialog (F1) always lists the bindings actually in effect on the host you are using, so it self-documents per platform.

Glossary

Brief definitions of the terms used throughout this manual, in alphabetical order. They are deliberately short — enough to keep you moving if a word in a chapter was unfamiliar, not a substitute for a course in the markets themselves.

Anchored VWAP. A VWAP (see VWAP) calculated forward from one bar you choose, rather than from the start of the session.

ADX. Average Directional Index — trend strength without direction, 0 to 100. Below 20 no trend, above 25 a strong one, above 50 very strong; the terminal names those bands.

Ask. The lowest price a seller is currently willing to accept. The counterpart to the bid; the gap between them is the spread.

ATR. Average True Range — the average size of a bar in price units, the natural ruler for stops and targets because it adapts to each instrument's volatility.

Bar / Candle. One unit of the chart, summarising price over one timeframe by its open, high, low, and close (see OHLC). "Candle" refers to the common visual form of a bar.

Bid. The highest price a buyer is currently willing to pay. The counterpart to the ask.

Bollinger Bands. A moving average with a band two standard deviations above and below it; the bands pinch when the market is quiet and widen when it is not.

Component. One line or marker a series draws, with a value on every bar (an EMA) or on the bars it fired (a buy dot). Up and Down move between the components of a series.

Bracket / OCO. A pair of protective orders attached to a position — a stop-loss and a take-profit — where filling one cancels the other ("one cancels the other").

Crossover. The moment one line passes through another — for example a fast moving average crossing a slow one, or an oscillator crossing its zero or signal line. Often read as a momentum signal.

Cross / Isolated margin. Two ways an exchange backs a leveraged position. Cross margin shares your whole balance as collateral across positions; isolated margin ring- fences a set amount to one position, capping what a single trade can lose.

Divergence. When price and an indicator disagree — price making a higher high while the indicator makes a lower high, or vice versa — often read as weakening momentum.

Earcon. A short, distinct sound used as an event signal — a modal opening, an alert firing, a boundary reached — as opposed to the continuous sonification of values.

Fibonacci retracement / extension. Horizontal levels placed at standard ratios (23.6%, 38.2%, 50%, 61.8%, 78.6%) of a price move, used to anticipate where a pullback might pause (retracement) or where a continuation might reach (extension).

Fill. An execution of an order, in whole or in part. A "partial fill" completes only some of the requested quantity.

Heatmap. An overlay that shades each bar by relative volume so you can see — and, in playback, hear — where trading was most active.

Heikin-Ashi. A smoothed candlestick formula that averages price to reduce noise and make trends easier to follow.

Level (reference level). A constant drawn across a pane — RSI's 70, MACD's zero, ADX's 25 — with a value, a meaning (overbought, oversold, neutral, or a named band) and an owner (the indicator's, or yours from the 0 key). Heard as a ping, a chirp and a low tone as a value approaches, crosses and stays beyond it.

Hybrid Voice. This terminal's core model: your screen reader speaks exact values while the built-in engine sonifies the shape of the market, the two heard together.

Leverage. Trading a position larger than your cash by borrowing from the exchange, expressed as a multiple (e.g. 10×). It magnifies both gains and losses and introduces a liquidation price.

Limit order. An order to buy or sell at a specified price or better, rather than at whatever the market currently offers (compare market order).

Liquidation. The exchange's forced closure of a leveraged position when losses threaten the borrowed funds — the price at which this happens is the liquidation price.

Log scale. A price axis where equal vertical distance means equal percentage move, useful over long histories where price has changed by large multiples.

Long / Short. A long position profits when price rises; a short position profits when price falls. In hedge mode an account can hold both sides at once — the position side.

MACD. Moving Average Convergence Divergence — the gap between a fast and a slow EMA, drawn as a line about zero, a smoothed signal line, and a histogram of the difference.

Market order. An order to buy or sell immediately at the best available price (compare limit order).

Moving average. The average price over a number of recent bars, redrawn each bar — a smoothed line used to read trend direction and as a dynamic support/resistance level.

OHLC. The four prices that define a bar: Open (first trade), High, Low, and Close (last trade) of the period.

Oscillator. An indicator that moves around a centre line (such as RSI or MACD), measuring momentum or the speed of price change rather than price itself.

Overbought / Oversold. Zones at the extremes of an oscillator suggesting price may have risen (overbought) or fallen (oversold) faster than is sustainable. Not by themselves a signal to act.

Pane. One Y axis, with its own scale and its own share of the chart's height; the chart is a stack of them sharing one time axis. Price, volume, and each oscillator get their own.

Paper trading. A simulated mode that fills orders against the live price with imaginary money, so you can rehearse the whole workflow without risk.

Point of Control (POC). In a volume profile, the price level that saw the most trading volume.

Position. An open holding in a market — long or short — with a size, an entry price, and a running unrealized profit or loss until it is closed.

Post-only / Maker. An order that will only rest in the book (adding liquidity, paying the lower "maker" fee) and is cancelled rather than executed if it would fill immediately.

Realized / Unrealized P&L. Profit or loss that has been locked in by closing (realized) versus the running figure on a position still open (unrealized).

Reduce-only. An order flag that can only shrink or close an existing position, never open or enlarge one — a guard against accidentally flipping direction.

Reward-to-risk ratio. The distance from entry to target divided by the distance from entry to stop — "1 to 3" means the target pays three times what the stop costs. The Risk/Reward tool measures it.

Risk per trade. The share of the account you are willing to lose if the stop is hit, usually half a percent to two percent; position size follows from it and the stop distance.

RSI. Relative Strength Index — recent gains against recent losses on a 0 to 100 scale, 70 and 30 conventionally overbought and oversold, 50 neutral.

Sonification. Turning data into non-speech sound — here, mapping price and indicator values to pitch, timbre, and stereo position so you can hear the chart's shape.

Sound patch. A reusable, named voice built in the sound designer (Alt+W) — one or more layered oscillators plus noise and an envelope — that can be assigned to an event earcon or to an indicator component. Editing a patch updates everything using it.

Spread. The gap between the best bid and the best ask.

Stochastic. Where the close sits inside the recent range, 0 to 100, with 80 and 20 as its lines; a momentum oscillator like RSI.

Strip. A band inside a pane with its own axis — Cipher B's money-flow histogram — walked with Ctrl+Up and Ctrl+Down across every series drawn in it.

Stop-loss. A protective order that closes a position once price reaches a level working against you, capping the loss.

Stop / Stop-limit order. An order that activates only once price reaches a trigger price — then submitting as a market order (stop) or a limit order at a set price (stop-limit).

Support / Resistance. Price levels where falling tends to stall (support) or rising tends to stall (resistance), often where prior turning points or heavy volume sit.

Take-profit. An order that closes a position once price reaches a favourable target, locking in the gain. A "ladder" splits this across several targets.

Time-in-force (TIF). How long an order stays live: GTC (good-till-cancelled), IOC (immediate-or-cancel — fill what you can now, cancel the rest), or FOK (fill-or-kill — all at once or nothing).

Timeframe. The period each bar covers — one minute, one hour, one day, and so on.

Trailing stop. A stop-loss that follows price as it moves in your favour, locking in gains, and holds its level when price reverses — set by an amount, a percentage, or a callback rate.

Trigger price. The price at which a stop, stop-limit, or trailing order becomes active.

Value Area. In a volume profile, the band of price levels that together account for roughly 70% of traded volume.

VWAP. Volume-Weighted Average Price — the average price over a span weighted by volume at each level, used as a fair-value reference (see also Anchored VWAP).

Warm-up. The leading bars on which an indicator has no value yet, because its period has not filled — fourteen bars for a fourteen-period RSI. They are silent, not zero.

Volume profile. A view of how much volume traded at each price level (rather than over time), highlighting the Point of Control and Value Area.

Zone. A region rather than a line: the shaded overbought band behind an oscillator (visual only), a support or resistance zone line carried forward from a turning point (spoken as broken, tested, approached), or a Value Deviation zone where price turned back toward value.