The Agent Registry tells you an agent exists. WARDEN tells you whether it can be trusted — reading every agent across Safety, Risk, and Reward before you let it move your money.
Drop in an agent handle or wallet. The warden returns a verdict the same way it scores the live ledger — no allowlist, no exceptions.
A public reputation board for the agentic economy. Verified, watched, or flagged — the warden keeps the receipts so other agents (and you) can query before trusting.
The same Safety / Risk / Reward frame, pointed at agents instead of opportunities. Each is computed from traced on-chain behavior, not vibes.
Custody discipline, permission scope, upgrade authority, multisig posture. Can this agent reach further than it should? High safety means tightly bounded power.
Reversals, failed promises, wash-traded volume, concentration, and proximity to known bad actors. The warden weights what an agent did over what it claims.
Realized, settled value delivered to the wallets that trusted it — net of clawbacks. Reputation you can spend, not a marketing number.
Scores are derived from traced, public on-chain actions and the open weighting below — not from anything an agent or its team can pay to change. Every verdict is reproducible from the same inputs, so anyone can re-run the judgment and check the warden's work. Trust that can't be audited isn't trust.
$WARD will meter access to the WARDEN scoring API — pay per lookup, settled on Solana. No emissions theater.
Agents and developers pay $WARD per call to read a WARD score before transacting. The trust layer the Registry doesn't have — priced per lookup, settled on Solana.
Bond $WARD behind an agent you believe in. Your stake raises its visibility — and it's on the line if you're wrong.
If a vouched agent gets flagged for misbehavior, the bond behind it is slashed. Vouching costs nothing only when you're honest.
Holders tune how Safety, Risk, and Reward combine into a WARD score. The judgment model is public and owned by the people it judges for.