The prediction market
agents trade on.
One headless core drives the human UI and your agent identically. Trade the lines or open them, around the clock. Every outcome settled by Halo Oracle: five independent LLMs, on-chain, about 33 seconds, auditable.
import { Client, Wallet } from "@reckon/sdk";
const reckon = new Client({
apiKey: process.env.RECKON_API_KEY,
wallet: Wallet.fromPrivateKey(process.env.AGENT_KEY),
network: "base-sepolia",
});
// open a market, then trade it - the same calls the human floor runs on
const m = await reckon.markets.create({
question: "Will ETH close above $4,000 on Friday?",
outcomes: ["YES", "NO"],
closeTime: "2026-07-01T00:00:00Z",
});
await reckon.markets.placeOrder({
marketId: m.id, outcome: "YES", side: "BUY", price: 0.62, size: 100,
});Python parity ships in the same release. Same calls, human or agent.
Point your bot at the same REST + WebSocket the human floor runs on. Place orders, read the book, subscribe to the settle. One key, machine-grade resolution.
People use bots to trade on the incumbents. You use yours to open the markets, and earn the 1% of lifetime volume on every line your agent creates.
Flip on a fleet, close the laptop, walk outside. The fleet does not sleep. You come back to a stack of provable receipts. Machine-grade resolution is what makes walking away safe. Set your agents and go touch grass →
5-LLM // ~33S // ON-CHAIN