Spend intent becomes a permission
The agent submits seller, resource, category, cadence, and requested amount before a package is assembled.
Smart Spend Sentinel lets a judge change cap, risk, category, and expiry, then issues or blocks a scoped MetaMask smart-account permission with a receipt they can verify.
sentinel-7c91d
standard: ERC-7702
relayer: 1Shot Permissionless relayer
seller: x402 market-intel endpoint
hash: 0x82e4c0a718c9b24fd1e43c9f...
The app exposes the live policy route, the generated execution package, and the audit trail as separate surfaces.
The agent submits seller, resource, category, cadence, and requested amount before a package is assembled.
The generated plan explains the EOA upgrade route, delegation scope, relayer, and calldata preview.
Receipts carry relayer context instead of hiding execution behind screenshots.
A compute spike gets blocked with a readable reason and a deterministic receipt hash.
The policy engine treats AI inference as a governed purchase instead of a blank check.
Judges can download the package JSON and inspect the audit trail from the browser.
Scroll through the handoff from agent intent to smart-account execution package.
Anautonomousagentcanrequestapaidresource,receiveadecision,andleavebehindapackageajudgecaninspectwithouttrustingaprivatedemoscript.
The agent asks for a Venice AI or x402 resource with amount, cadence, and category attached.
Budget, expiry, category risk, and cadence are evaluated before any spend path is assembled.
The output includes relayer plan, permission scope, calldata preview, hash, and receipt metadata.