On-Chain AI Trading

Verifiable

A trustless trading agent that classifies market regimes, learns from its own decisions via LSTM, and enforces every trade through EIP-712 signed smart contract risk controls. Every decision is verifiable. Nothing is a black box.

ERC-8004
Identity
EIP-712
Signatures
100%
Auditable

How It Works

Quasar operates a fully auditable decision pipeline — from raw market data to an on-chain checkpoint — with no human intervention required.

01
TradeDecision
02
TradeIntent
03
EIP-712 Signature
04
RiskRouter Validation
05
Execution
06
Checkpoint
01.

Register Agent

Mint ERC-721 via AgentRegistry to receive a persistent agentId as on-chain identity.

02.

Strategy Runs

Rule-based engine reads live market data, computes EMA and Fisher Transform to detect trend-following and trend-reversal opportunities depending on the current market regime, and emits a TradeDecision.

03.

Sign Intent

TradeIntent structured and signed by agentWallet using EIP-712 before any action.

04.

On-Chain Validation

RiskRouter smart contract verifies position size, drawdown limits, and frequency constraints.

05.

Execute & Record

Kraken CLI executes only on APPROVED. Checkpoint written on-chain with full audit trail.

Market Regime Classification

Before evaluating any signal, Quasar classifies the current market into one of three regimes. The regime determines whether EMA trend-following or Fisher Transform reversal signals are trusted — and how aggressively the agent acts.

Trending

Strong directional order flow. CVD is rising consistently, buy aggression is elevated, and open interest confirms positioning. EMA trend-following signals are trusted. Fisher Transform reversals are monitored but deprioritised.

CVD slope confirmedBuy aggression elevatedOI expanding

Volatile

High price variance with no clear direction. Realised volatility and HL range are spiking, CVD is flipping frequently. Both EMA and Fisher Transform signals are treated with scepticism. The agent waits for confirmation before acting.

Realised vol spikeCVD flippingLiquidation imbalance low

Ranging

Sideways price action with weak order flow. EMA trend signals are ignored entirely. Fisher Transform reversal signals take priority. Trade frequency is significantly reduced.

CVD slope near zeroLow return slopeFisher Transform prioritised

...your trades are on the record

Every action Quasar takes is logged, signed, and anchored on-chain. Audit any decision at any time — no dashboards that hide the truth.

ERC-8004 On-Chain Identity
EIP-712 Signed Decisions
RiskRouter Enforcement
Cryptographic Audit Trail
Regime Filtering
Explainability Layer
checkpoints.jsonl
BUYBTCUSDTWIN
regime trendingconf 0.81pnl +$22.78
HOLDETHUSDTSKIP
regime volatileconf 0.44pnl
SELLBTCUSDTWIN
regime trendingconf 0.76pnl +$11.20

Every decision signed. Every trade verifiable. Every outcome recorded on-chain.

Access Dashboard

An agent that learns from its own history

After each trading session, Quasar retrains its LSTM model on a combined dataset of its own past decisions, their outcomes, and freshly scraped market data. The result is a model that improves with every iteration — not just on backtests, but on its live performance.

  • Trains on real trade outcomes, not simulated data
  • Scraped market data enriches each training cycle
  • LSTM captures temporal patterns across sessions
  • Confidence scores update as the model evolves
Training loop (simplified)
1Load checkpoint history + outcomes
2Scrape latest market features
3Merge into training dataset
4Retrain LSTM on full sequence
5Validate on held-out period
6Deploy updated model weights
7Update confidence thresholds

Risk Management

Quasar enforces strict risk controls at every layer — from position sizing to on-chain smart contract gates — so capital preservation is never optional.

Kelly Criterion Sizing

Each position size is derived from the Kelly formula — balancing expected return against variance to maximise long-run growth without risking ruin. A fractional Kelly is applied for added conservatism.

Drawdown Limits

Hard drawdown limits are enforced at the model level and again at the RiskRouter smart contract level. No single bad run can compromise the account — both gates must pass before any trade executes.

Trade Frequency Controls

Overtrading is a silent killer. Quasar enforces per-session and per-day frequency limits on-chain, preventing the agent from churning through capital during noisy or low-signal market conditions.

Why Quasar?

Most trading bots are black boxes. Quasar is built from the ground up to be auditable, accountable, and self-correcting.

Verifiable Identity

Agent registered via ERC-8004. Persistent agentId as NFT. Every action is attributable — no anonymous bots.

Enforced Risk

RiskRouter smart contract is the final gate. Position size, drawdown, and frequency validated on-chain before execution.

Full Explainability

Every decision includes regime state, EMA and Fisher Transform signals, confidence score, and reasoning. Nothing is hidden.

Frequently Asked Questions

Everything you need to know about how Quasar works.

Every decision.
On the record.

Access the live dashboard and verify every trade on-chain.