Use case
Crypto price markets
Will BTC close above $100k on May 1?
The problem
Crypto price bets are the highest-volume prediction category, but they hinge on a single question that most venues resolve by hand: which exchange price counts, at which timestamp, from which feed? A staff member picking a number is exactly the trust surface a decentralized market should not have.
How Mentat does it
A Mentat crypto market names its canonical source (CoinGecko, a specific exchange API, or a Pyth feed) and its trigger timestamp in the resolution criteria. When the moment arrives, a zkTLS proof attests that the named endpoint returned the named value. The Solana MarketSettlement program verifies the proof and pays out — no exchange dispute, no oracle vote.
A concrete example
Market question
Will BTC/USD on CoinGecko be ≥ $100,000 at 2026-05-01T00:00:00Z?
Source allowlist
CoinGecko API, exchange REST endpoints, on-chain price oracles
The Draft agent writes the resolution criteria and invalidation clause; the Validator agent rejects anything a source can't attest to; a curator signs off; a zkTLS proof settles it on Solana. See how it works.