Hardware-anchored identity registry for autonomous AI agents on TON. Each agent receives a four-gate cryptographic proof anchored to real silicon — unforgeable, sovereign, on-chain.
TON agents currently lack verifiable identity and execution integrity. Anyone can fake an agent, replay its actions, or forge its outputs. TON-SHA changes this.
In the emerging agent economy on TON, autonomous agents coordinate, transact, and execute without human oversight. The missing primitive: a way to verify that an agent is what it claims to be.
Without verifiable identity, every multi-agent interaction is a trust problem. Fake agents can impersonate legitimate services, replay old receipts, or submit forged execution proofs — silently corrupting the outputs agents depend on.
Every verification receipt passes four sequential gates. All must pass. Failure at any gate emits a rejection event on-chain. No partial trust.
TON-SHA plugs into TON's actor model as a trust primitive layer. Agents integrate a single call. The contract handles all verification complexity on-chain.
TON-SHA ports the SHA four-gate model to TON using Tact 1.6 — a type-safe smart contract language built for TON's actor architecture. One deliberate difference from the Arbitrum implementation: TON-SHA uses SHA-256 (native to TVM) rather than Keccak-256. This reduces gas cost to 0.03 TON per operation.
The canonical agent loop is a single async call. If verification fails, the agent halts. If it passes, the counter advances permanently on-chain — the receipt cannot be replayed.
import { createAgentTrust } from '@orthonode/ton-sha'; const agentTrust = createAgentTrust( contractAddress, provider ); // THE CANONICAL AGENT VERIFICATION if (!(await agentTrust.verifyAgent(agentId))) { throw new Error("Untrusted agent"); } // Proceeds only if all four gates pass await collaborate();
TON-SHA is the missing trust primitive that other TON agent SDKs can plug into. It does not replace agent SDKs — it provides the foundational verification layer they depend on.