Loops HouseSignatory

NEAR HACKATHON
Links
Team
1 member- BAOwner
Balamurughan G
Overview
Signatory is Agentic Business at machine speeds with hybrid officer–agent org charts and rigorous AI governance, enforced on NEAR.
Two sample companies with synthetic data, Tommy Hilfiger Europe (buyer) and Jupiter Knitting (seller), trade through AI agents. Each agent acts for a named human officer. Its authority is a vLEI credential chain: GLEIF → QVI → company (LEI) → officer with an OOR role credential → AI agent → helper sub-agents, all on KERI key event logs.
One run, live on NEAR testnet:
Tommy’s buyer agent sends an order. Jupiter verifies the buyer agent live: key logs plus the officer’s OOR credential presented to a verifier (Sally). Jupiter’s seller agent asks its sub-agents (inventory, logistics, treasury) and agrees a price. Tommy’s treasury agent checks the agent’s mandate, stored in our NEAR contract. Jupiter issues the invoice as a signed vLEI credential, handed over with IPEX. A registrar recomputes its fingerprint and registers it on NEAR. A tampered invoice never reaches the chain. The buyer agent pays USDC through the contract: within its limit it pays, above it the Chief Procurement Officer must approve, and if revoked the payment is refused and refunded in the same transaction.
Built for this hackathon: the NEAR mandate contract (chainaim-signatory-v2.testnet), signer, bridge, runner, agent swarm and negotiation, AP2 signing, and the escalation console.
Why Signatory is different: other agents get a wallet. Ours get a boss, a legal identity, and a limit NEAR enforces.
Proof: the README links real testnet transactions, and npm test runs 141 tests with no keys.