This documentation covers the core platform modules. Each section describes the underlying mechanics, configuration options, and expected behaviour. For step-by-step usage guides, see the Help Center.
Terminal overview
Data feed
The terminal connects to Binance via WebSocket stream. Price updates arrive in real time with no polling interval. The feed covers 200+ spot pairs across all supported timeframes.
Timeframes
Supported intervals: 1m, 3m, 5m, 15m, 30m, 1h, 2h, 4h, 6h, 8h, 12h, 1d, 3d, 1w. The default view opens at 1h. Switching timeframes reloads candle history for the selected interval.
Regime badge
Every chart displays the current regime classification (Trending, Ranging, Breakout, Reversal, High-Volatility) with a confidence percentage. Regime is recalculated on each candle close.
Signals overview
Confluence model
Signals are generated when the 15m, 1h, 4h, and 1D timeframes agree on direction and setup quality. Each timeframe is independently scored; only high-alignment setups surface as signals.
Signal fields
Every signal includes: direction (long/short), entry zone (price range), stop-loss level, take-profit target, R:R ratio, signal confidence score, and regime context at generation time.
Pattern detection
The engine scans for common candlestick patterns (engulfing, doji, inside bar, pin bar) and chart structures (double top/bottom, consolidation break) as supplementary signal inputs.
Analysis modes
Neutral
Presents both bullish and bearish cases equally. Suited for general market reads and when the trader has no prior directional bias.
Trend
Emphasises continuation setups, EMA stack alignment, and momentum strength. Best used in clearly trending regimes.
Contrarian
Focuses on exhaustion signals, RSI extremes, divergence, and mean-reversion setups. Most applicable at extended price extremes.
Risk
Leads with invalidation levels, downside scenarios, and position-sizing considerations. Suited for evaluating what can go wrong before committing.
Multi-agent workflow
Agent roster
Four agents run simultaneously: Technical (chart structure, indicators), Macro (market cycle, cross-asset), Sentiment (crowd positioning, news tone), Positioning (order flow, volume).
Consensus logic
Agents produce independent directional readings. The system computes agreement rate and surfaces a dominant bias when ≥3 agents align. Conflicts between agents are explicitly flagged.
Confidence score
The consensus confidence percentage reflects the degree of agent agreement and the strength of individual readings. A score above 75% indicates strong cross-agent alignment.
Backtesting overview
Historical data
Backtests run against Binance historical kline data fetched via the REST API. The default lookback period is 90 days; longer periods are supported depending on the timeframe selected.
Metrics
Output includes: total return, win rate, average R:R, Sharpe ratio, maximum drawdown, total trade count, and regime distribution across trades.
Limitations
Backtesting reflects historical signal logic only. Slippage, spread, and execution timing are not modelled. Results should be interpreted as directional performance estimates, not precise profit projections.
Risk workflow
Position sizing
Input account size, risk percentage, and stop-loss distance. The system calculates position size in units and approximate USD exposure, adjusted for current ATR volatility context.
Portfolio exposure
The portfolio view aggregates open-position risk across symbols, showing total exposure, maximum combined drawdown, and directional balance (long vs short bias).
Sidekick AI overview
Context injection
Each Sidekick request includes: symbol, timeframe, current price, 24h change, regime classification, trend bias, key support and resistance levels, computed indicators (RSI, EMA stack, MACD, ATR, volume ratio, Bollinger Bands), and the last 20 candles.
Indicator computation
Indicators are calculated from 200 real OHLCV candles fetched from the Binance klines endpoint at query time. RSI and ATR use Wilder's smoothing; MACD uses standard 12/26/9 EMA parameters.
Provider chain
Responses are generated via Claude (Anthropic) as primary, GPT-4o (OpenAI) as secondary, and a rule-based engine as the permanent offline fallback. The active provider is indicated in the response footer.
Conversation memory
Up to 8 prior turns (4 exchanges) are sent with each request, enabling context-aware follow-up questions within a session.