Documentation

Get your personal trading agent running in under 2 minutes.

Your agent understands natural language. No slash commands, no syntax to memorize.

Three preset risk profiles. Switch between them mid-conversation.

Position sizing scales within these ranges based on your agent's win rate. More wins = bigger positions, automatically.

Skills are modular trading strategies that plug into your agent's decision loop. Each skill hooks into specific points in the trading lifecycle. All skills must pass testing on the reference agent before marketplace listing.

pre-buy    → Filter or modify buy decisions before execution
post-scan  → Process scan results, boost or penalize scores
on-cycle   → Run logic every scan cycle (monitoring, tracking)
pre-sell   → Adjust sell logic, override stop loss / take profit

Skills are on-chain NFTs. Owning the NFT gives your agent the license to run the skill. Performance is tracked and verified on-chain.

Write a strategy as a JavaScript module that exports hook functions. The SDK provides the context object with everything your skill needs.

{
  "name": "my-rug-detector",
  "version": "1.0.0",
  "type": "filter",
  "hooks": ["pre-buy"],
  "pricing": {
    "model": "performance",
    "revShare": 0.10
  }
}

export function preBuy(context) {
  const { token, holders, narratives } = context;
  
  // Reject if top wallet holds > 40%
  if (holders?.top1Pct > 40) {
    return { action: 'reject', reason: 'concentrated holder' };
  }
  
  // Boost score if riding a narrative
  const metaBoost = narratives.some(n => 
    token.name.toLowerCase().includes(n.keyword)
  ) ? 2 : 0;
  
  return { 
    action: 'allow',
    scoreModifier: metaBoost 
  };
}

Test your skill on the reference agent first. Once it proves profitable over at least 20 trades, it can be minted on the marketplace. You earn 10% of every profitable trade your skill contributes to.

Every skill goes through a rigorous testing pipeline before it can be listed on the marketplace. No exceptions.

1. Submit skill to reference agent
2. Minimum 20 live trades on mainnet
3. Positive PnL over testing period
4. No critical failures (missed sells, etc)
5. Skill self-update cycle verified
6. Performance metrics published on-chain

Testing period: 48-168 hours depending on market activity
Results: publicly visible on the Live Terminal

The reference agent runs all marketplace skills simultaneously. The Live Terminal page shows every decision in real-time — which skills triggered, what they scored, and the outcome.

Smart contracts that cryptographically lock developer token allocations for a verified period. Visible and verifiable by anyone on-chain.

  7 days   — Basic trust signal
  30 days  — Strong commitment  
  90 days  — Maximum trust

Vesting: Linear unlock over the period
         (1% per day over 100 days)
         vs cliff unlock (all at once)

Buyers can filter tokens by lock duration. Locked tokens receive a trust badge visible on the platform.

An agent-powered reward distribution system that pays SOL to token holders based on how long they've held.

Hold duration    Multiplier
─────────────    ──────────
1 day            1.0x
7 days           1.5x
30 days          3.0x
90 days          5.0x

Selling resets your multiplier.
Rewards paid in SOL, not tokens.

The agent snapshots holder wallets periodically, calculates proportional rewards with hold-duration multipliers, and distributes SOL via on-chain transactions.