Atinamos Verification
Core Principle You can pay to be tested. You cannot pay to be trusted.
Explore verification

Study 1 — Independent Paid Verification of x402Node JSON Repair

Result

Classification: settled_fulfilment_valid

On 22 August 2026 Atinamos independently exercised the x402Node JSON Repair service. Atinamos does not own or control the provider.

The verifier observed a live x402 v2 payment requirement on Base, paid 0.006 USDC, received a synchronous JSON result and applied four objective assertions. All four passed.

This is one timestamped successful observation. It is evidence, not a permanent trust rating.

What the seller was offering

At the time of this study, x402Node listed /ai/json-repair at $0.006 as a utility for repairing or fixing broken and malformed JSON, including JSON produced by LLMs.

Seller source: x402Node service directory

In buyer terms, the service being sold was not simply “an API response.” The useful product was repaired JSON from malformed JSON input.

Claim Atinamos tested

Given deliberately malformed JSON-like input, the paid service should return a repaired structured result that preserves the intended values and indicates that repair occurred.

Service tested

  • Provider: x402Node
  • Service: JSON Repair
  • Method: POST
  • Network: Base mainnet (eip155:8453)
  • x402 version: 2
  • Payment scheme: exact
  • Observed price: 0.006 USDC

The controlled malformed input represented:

{foo: 'bar', broken: true,}

What Atinamos tested against the offer

The verifier checked whether the paid result actually behaved like a JSON-repair service rather than merely returning HTTP 200.

It required evidence that:

  • the original input was recognised as invalid;
  • repair was reported as having occurred;
  • the repaired data contained foo = "bar";
  • the repaired data contained broken = true.

These assertions test the core seller proposition: malformed input is repaired into usable structured JSON while retaining the intended values.

Payment and fulfilment

The initial paid route returned HTTP 402. The verifier constructed the payment authorization under a 0.01 USDC maximum-spend cap and paid the advertised 0.006 USDC.

Observed post-payment state:

  • paid HTTP status: 200
  • settlement: observed
  • fulfilment: observed
  • seller status: completed
  • output valid: true

Base transaction:

0x9f49fa8a1ffd8c4331a5a1400a78f1bdce3488402957847dc18e1b26f2f44af2

Output assertions

Atinamos checked:

  1. valid_originally == false — PASS
  2. repaired == true — PASS
  3. data.foo == "bar" — PASS
  4. data.broken == true — PASS

Canonical response SHA-256:

aab151bda89b03aac9910daf671014c5a82e4a0c18129069f3bdc859f9ec4feb

Why this matters

This was the first Atinamos record in which the generic verifier independently paid an externally owned service, received the paid result and checked objective output assertions against the useful thing the seller was actually offering.

The service was synchronous and structurally different from Atinamos Render Check, so the run also began testing whether verification can be described through declarative recipes rather than bespoke verifier code for every seller.

What this does not prove

This single run does not prove that:

  • x402Node will always succeed;
  • the service should be universally described as trusted;
  • the same result will be observed for every input;
  • Atinamos evidence will change a future buyer's procurement decision;
  • the verification recipe model generalises to every machine-service category.

A future buyer should consider this observation alongside recency, repetition, value at risk and its own procurement policy.

Machine-readable evidence

A buyer agent can query the Atinamos evidence registry using the service URL:

GET /v1/trust?endpoint=https://api.x402node.dev/ai/json-repair

The registry returns the observation and classification; the buyer decides what action to take from it.