Overview

Crypto Signal AI monitors configured markets, detects strategy setups, explains them in plain language, applies risk checks, and delivers transparent alerts — so you always know why a signal appeared and whether it passed approval.

Research and education only.The app does not place trades and does not provide financial advice. Alerts are decision support — review every signal manually before acting.

What You Get

When a setup is detected, you receive a structured alert — not a vague tip. Each message includes the symbol and timeframe, long or short direction, strategy name, entry/stop/target prices, reward-to-risk, signal strength, AI explanation, strategy trigger reasons, conflict notes, and why risk approved or rejected the signal.

Clear trade levelsEntry, stop loss, and take profit are defined upfront — not implied after the move.

Readable contextExplanations describe regime alignment, indicator state, and anything the scorer flagged as conflicting.

Honest rejectionsSignals that fail risk checks are still logged and can alert — so you see filtered setups too, not just winners.

See a full field-by-field guide in Signal Workflow.

Design Principles

The platform is intentionally risk-first and rule-based first. These four principles shape every layer — from how signals are generated to how the product evolves over time.

Rule-Based First

Signals originate from deterministic strategies that are inspectable, testable, and backtestable — not opaque buy/sell calls.

Today

EMA RSI Breakout — trend-following
Mean Reversion — range reversals
Volatility Breakout — compression breakouts

Planned

Trend pullback, z-score mean reversion, funding-rate divergence, and more

Learn more in Strategies.

AI as Analyst

AI scores setup quality, writes plain-English explanations, and flags conflicts. It does not bypass risk controls or place trades.

Today

Signal quality scoring (0–1) with pass/reject decision
Plain-English explanations and conflict notes
OpenAI when configured; deterministic rule engine as fallback

Planned

Richer regime context and historical performance in prompts
Separate ML trade-quality prediction from explanation

Learn more in AI scoring.

Risk Before Profit

Every final signal passes hard risk checks before it is marked approved. Risk management has final authority over AI scoring.

Today

Daily approved-signal cap
AI rejection enforcement
Minimum reward-to-risk (1.0)
Maximum stop-distance check
Position size suggestion with human-readable risk_reasons

Planned

Drawdown circuit breaker, consecutive-loss halt, portfolio exposure limits

Learn more in Risk Management.

Data-Driven Evolution

Signals and outcomes are stored so performance can be measured over time — the foundation for analytics, ML filtering, and strategy refinement.

Today

Full signal audit records in PostgreSQL
Outcome tracking for take profit, stop loss, expiry, and pending
Backtesting with expanded metrics (stdout today)

Planned

Strategy performance by regime, ML quality filter, strategy retirement rules

Learn more in Outcome Tracking.

3Implemented strategies

4Outcome states tracked

0Live execution paths

See the full implementation status table on the Product Roadmap. Configure markets and alerts on Configuration.

Platform Architecture

Data flows in one direction through fixed layers. Strategies and backtesting branch from enriched market data; outcome tracking runs after signals are stored. Nothing skips AI scoring or risk checks.

Exchange APIs
  → Market Data (CCXT OHLCV)
  → Indicators + Regime Analysis
  → Strategy Engine
  → Raw Signals
  → AI / Rule-Based Evaluation
  → Risk Manager + Position Size Suggestion
  → PostgreSQL
  → Telegram / Discord Alerts
  → Outcome Tracking
  → Backtesting (parallel research path)
  → Future ML Filter (planned)

Layer	Role	Status
Market Data	Fetches OHLCV candles from the configured exchange for each symbol and timeframe.	Available
Indicators + Regime	EMA, RSI, MACD, ATR, Bollinger Bands, ADX, and regime classification with confidence scores.	Available
Strategy Engine	Registry-based pluggable strategies; each emits entry, stop, target, and trigger reasons.	Available
AI Evaluation	Scores and explains each raw signal; adds conflict notes including cross-strategy disagreements.	Available
Risk Manager	Applies daily caps, min R:R, stop distance, and AI rejection rules; suggests position size.	Available
Persistence	PostgreSQL stores signals and outcomes; Alembic manages schema migrations.	Available
Alerts	Telegram and Discord delivery for approved and rejected final signals.	Available
Outcome Tracking	Resolves approved signals against future candles — TP, SL, expired, or pending.	Available
Backtesting	Historical simulation with Sharpe, Sortino, Calmar, profit factor, and more. Not persisted yet.	Available
ML Filter	Machine-learning quality scoring from historical outcomes. Planned Phase 3.	Planned

For step-by-step detail on each stage — including sample JSON and alert examples — see Signal Workflow.

Strategies

Three complementary rule-based strategies cover trend continuation, range reversals, and volatility breakouts. Each publishes the exact conditions it checks — no proprietary mystery logic.

EMA RSI Breakout

Trend-following setup using EMA alignment, RSI recovery or rollover, and MACD confirmation.

Mean Reversion

Range-oriented reversal setup using RSI and Bollinger Band context for stretched prices.

Volatility Breakout

Compression-to-expansion setup using Bollinger Band width and recent range breaks.

Full setup rules, regime filters, and exit logic are documented on the Strategies page.

Signal Pipeline

Every alert follows the same path — from raw market data to a reviewed final signal. Nothing skips scoring or risk checks.

1. Market scan
Live exchange candles are loaded for configured symbols and timeframes, then enriched with technical indicators and regime classification.
2. Strategy match
Each enabled strategy evaluates the same snapshot and may produce a signal with entry, stop loss, take profit, and trigger reasons.
3. Score & explain
AI or rule-based scoring rates setup quality, writes an explanation, and flags conflicts such as regime mismatch or weak volume.
4. Risk approval
Daily caps, minimum reward-to-risk, stop distance limits, and AI rejection rules are applied before a signal is marked approved or rejected.
5. Alert delivery
Final signals reach Telegram or Discord with trade levels, strength, explanation, sizing hint, and risk notes — approved or rejected.

Dive into each stage — including sample JSON and alert examples — on Signal Workflow. Risk rules and outcome resolution are covered separately on Risk Management and Outcome Tracking.

Explore Docs

ConfigurationEnvironment variables for exchange, alerts, risk, and outcome tracking.Signal WorkflowHow signals are built, what each field means, and how to read alerts.StrategiesExact entry rules, regime filters, and stop/target logic for every strategy.Risk ManagementApproval checks, position sizing, and what approved vs rejected means.Outcome TrackingHow signals are resolved against future price action over time.Product RoadmapImplementation status, phases, success metrics, and long-term vision.

Looking Ahead

The product is built in deliberate stages — from transparent alerts today toward validated research tools and, only after strict gates, optional live execution. That future direction is documented separately so the guides above stay focused on what you can use now.

Product RoadmapImplementation status, planned phases, success metrics, and long-term vision — including why live trading stays gated and opt-in.