Architecture

Designing Agent-First applications

A paradigm shift in application design. Learn how to interface your autonomous AI agents with structured, verifiable consensus planes.

The Three Pillars

1. Structured API Boundaries

Agents require typed data, not conversational interfaces. Every input claim and returned proof receipt is formatted as a strict, versioned JSON schema. This enables reliable code-level validation inside autonomous loops.

2. Idempotency & Replay Safety

In unreliable networks, agents will retry failed requests. By forcing a client-generated `idempotency_key` on all proof attempts, Verified OS guarantees that duplicate agent actions do not result in double charges on the transaction ledger.

3. Cryptographic Provenance

Proof-of-claim receipts are signed and stored with cryptographic signatures. Agents can retrieve, verify, and pass these receipts down the line, establishing a secure chain-of-custody for audited decisions.

Programmatic Verification

The following python example shows how an agent submits a claim to the verification plane:

import requests
import uuid

def verify_agent_claim(claim: str, evidence: str, token: str):
    response = requests.post(
        "https://api.verifiedos.ai/v1/verify",
        headers={"Authorization": f"Bearer {token}"},
        json={
            "claim_text": claim,
            "evidence_text": evidence,
            "idempotency_key": str(uuid.uuid4())
        }
    )
    response.raise_for_status()
    result = response.json()
    return result["public_result"] == "VERIFIED"