Prediction-market agents
that show their work.

01 · Research

Reads the market like an analyst.

An LLM agent studies prices, order books, trader track records and resolution rules. Its output is a strategy blueprint: what to trade, what triggers it, and the limits it must respect.

Has no access to any wallet key.

blueprint {
  trigger        WHALE_COPY
  targets        2 wallets
  max_position   $200
  abort_if_spread > 200 bps
}

02 · Signal

Waits for the moment.

Watchers stream live order books and whale fills. When a blueprint's trigger fires, the trade is sized against the bot's bankroll, scaled to how big the whale's bet was, and turned into an order intent.

intent {
  side        BUY YES
  notional    $75
  limit       0.42
  ttl         300s
}

03 · Execution

Checks first, then acts.

Four risk checks run before a key is touched. If all four pass, the order is signed in memory and sent straight to the venue's order book. If one fails, nothing is sent and the reason is logged.

risk_gate  ✓ cash  ✓ spread  ✓ expiry  ✓ exposure
order      fill-or-kill · EIP-712 signed

04 · Settlement

Closes the loop.

Winning positions are sold near $1 or redeemed when the market resolves. Fees are collected in batches, designed so bot wallets never need to hold gas.

position  resolved YES
exit      sold @ 0.995
fees      swept in batch