Autonomous Alpha Generation

Onchain Hedge Fund OS

Four specialized agents. One job each. Zero overlap. This is how institutional-grade DeFi strategies get built—debuggable, extensible, and fully autonomous.

Agent Architecture

Each agent has one job and one contract. No agent does two jobs. This keeps it debuggable and extensible.

[ Data Sources ]
Research Agent
→ Hypotheses
Portfolio Optimiser
Execution Agent
Performance Agent

Research Agent

Alpha Factory
Objective: Turn noisy information → structured trade hypotheses. Not "buy SOL". But: "SOL 7-day momentum + rising perp OI + social acceleration → long bias, 3–5 day horizon, medium confidence"

Data Sources (Phase 1)

  • cookie.fun → trending narratives, token velocity
  • Twitter (X) → curated accounts only
  • Price, Volume, Open Interest
  • Funding rates

Feature Extraction

  • Social momentum = Δ mentions / time
  • Price momentum = EMA(7) − EMA(21)
  • Funding skew = long vs short pressure
  • OI expansion = leverage entering/exiting

Hypothesis Output Schema

{
  "asset": "SOL",
  "bias": "LONG",
  "strategy": "Momentum",
  "time_horizon": "3-5 days",
  "signals": {
    "social_momentum": 0.82,
    "price_momentum": 0.74,
    "funding_rate": 0.12
  },
  "confidence": 0.68,
  "invalidated_if": ["Funding flips negative", "Price breaks 20D VWAP"]
}

Agent Pipeline

Raw signals → Feature extraction → Narrative clustering → Hypothesis generation → Confidence scoring

Portfolio Optimiser

Capital Brain
Objective: Convert hypotheses → capital weights. Start simple: Mean–Variance (Markowitz), volatility cap per asset, stablecoin buffer. Easy to debug, easy to upgrade to CVaR, Kelly fractions, or regime switching.

Inputs

  • Expected return → confidence × momentum
  • Covariance → rolling returns
  • Risk budget → user profile
  • Stablecoin buffer (20–40%)

Constraints

  • Σ weights = 1
  • w_i ≤ max_weight per asset
  • Stablecoins ≥ 20–40%
  • Volatility cap per position

Output Schema

{
  "timestamp": "2026-01-27",
  "weights": {
    "SOL": 0.25,
    "ETH": 0.20,
    "BTC": 0.15,
    "USDC": 0.40
  },
  "risk_metrics": {
    "expected_vol": 0.18,
    "expected_return": 0.32
  }
}

Optimisation Math

μ_i = confidence_i × momentum_i → max_w μᵀw − λ wᵀΣw

Execution Agent

Hyperliquid Integration
Objective: Turn weights → real trades with minimal slippage. Rules-based, deterministic. No LLM touches this. Execution Agent should be stateless + idempotent.

Execution Logic

  • Target weight − Current weight = Δ exposure
  • Split orders for size
  • Choose strategy based on market state
  • Execute with failure recovery

Strategy Mapping

  • High momentum → Market / Aggressive limit
  • Choppy → Grid orders
  • Mean reversion → VWAP / TWAP
  • Low liquidity → Scaled entry

State Management

  • Order state tracking
  • Position state monitoring
  • Failure recovery protocols
  • Testnet first, always

Hyperliquid Specifics

  • Use testnet for validation
  • Maintain order book awareness
  • Gas optimization
  • Slippage protection

Performance Agent

Truth Teller
Objective: Tell you what actually works. Track every strategy, score every hypothesis, feed learnings back into the system. This feedback loop is Nexxore's moat.

Core Metrics

  • Sharpe Ratio per strategy
  • Max Drawdown tracking
  • Win rate & profit factor
  • Volatility-adjusted PnL

Advanced Analytics

  • Alpha vs BTC benchmark
  • Strategy attribution
  • Hypothesis accuracy scoring
  • Signal decay analysis

Feedback Loop Example

// Strategy Performance
Momentum on SOL    → Sharpe 1.8  ✓ Increase allocation
Mean-rev on memes  → Sharpe 0.4  ✗ Reduce allocation

// Next Cycle Adjustments
{
  "action": "Increase confidence threshold for low-Sharpe strategies",
  "capital_shift": "Allocate less to strategies with Sharpe < 1.0"
}

Get Access to Agents

Unlock autonomous alpha generation with our agent system. Simple, transparent pricing.

Free Tier

Explorer

$0 / month

Get started with basic access to understand the system.

  • View agent architecture
  • Research Agent (read-only)
  • Daily market summaries
  • Portfolio Optimiser
  • Execution Agent
  • Performance Analytics
Full Access

Agent Pro

$19 / month

Full access to all four agents with autonomous execution capabilities.

  • Research Agent — Full hypothesis feed
  • Portfolio Optimiser — Custom weights
  • Execution Agent — Hyperliquid integration
  • Performance Agent — Full analytics
  • Real-time alerts & notifications
  • Priority support & Discord access

Pay with ETH on Ethereum mainnet. Access granted within 24 hours.

Pay with Ethereum

Send exactly $19 worth of ETH to unlock Agent Pro

Payment QR Code
Ethereum Address
0x905aCd442c7B3EF9BfEB0A3189f3686c1Cd0c697
$19 in ETH (≈ 0.004 ETH)

After payment, send your wallet address to access@nexxore.io for verification. Access granted within 24 hours.

Address copied!

Ready to Automate Your Alpha?

Join hundreds of traders using Nexxore's autonomous agent system.

Get Access Now Back to Home