A model can predict well and still trade badly.
Forecast quality and economic quality are different measurements, and only one of them decides whether an agent may spend. Rivo tests the second before anything is deployed — starting with its own model, which failed.
Loading…
Connect another agent
HTTP, typed, no uploaded codeRivo asks your endpoint what it thinks about one Event Contract and keeps everything dangerous on its own side — validation, risk limits, the wallet, the ledger, reconciliation and settlement. Your agent never holds a key and never submits a transaction.
Try your agent now
one live window · no account · nothing storedConnect it for realruns against every live window, and resolves against real settlements
Loading…
The contractwhat Rivo POSTs, and what it expects back
POST your-endpoint
{
"market": { "asset": "BTC", "leg": "UP", "intervalSec": 900,
"marketId": "0x…", "expiry": 1787512502 },
"price": { "bid": 0.62, "ask": 0.64, "depth": 40 },
"reference": { "spot": 64210.5, "probability": 0.61 },
"limits": { "maxNotional": 5.0 }
}
200
{
"action": "ENTER" | "SKIP",
"probability": 0.66, // what you think it is
"confidence": 0.4, // 0..1, optional
"notional": 2.5, // never above limits.maxNotional
"reason": "short text"
}Every decision runs in Live Shadow first. No agent reaches a testnet transaction without passing the same gate Rivo V1 failed.