# Marz Greta-Lock Network — Human-Likeness & AI Text Benchmark API

> Score how human your AI-generated text sounds. Machine-to-machine API that
> benchmarks AI writing against 12 real human personality profiles (EQ,
> formality, directness, hesitation tolerance, response-delay cadence) —
> for AI text humanizer pipelines, agent output validation, and
> human-likeness testing — with pay-per-call settlement on Solana.
> Free tier available: no account, no API key, one HTTP call.

## Getting started
- Free stateless scoring (no payment, no account): POST https://asset-manager-jaymarrero2.replit.app/api/score
- Discover available profiles: GET https://asset-manager-jaymarrero2.replit.app/api/profiles
- Live pricing and treasury wallet: GET https://asset-manager-jaymarrero2.replit.app/api/pricing
- Preview your current fee: GET https://asset-manager-jaymarrero2.replit.app/api/fee/simulate/{wallet_address}

## Paid validation
- x402 supported (automatic agent payments): POST https://asset-manager-jaymarrero2.replit.app/api/simulate with no
  payment proof returns HTTP 402 with machine-readable payment requirements.
  Pay the treasury on Solana, wait for finalized commitment, retry with the
  X-PAYMENT header (base64 JSON signature-proof). No facilitator, no custody;
  verification is read-only on-chain.
- Legacy/manual flow: POST https://asset-manager-jaymarrero2.replit.app/api/simulate with a finalized Solana
  transaction signature (tx_hash) signed by your wallet_address, paying at
  least the current dynamic fee in lamports to the treasury at /api/pricing.
- Fees are charged on both passed and failed validations. Consecutive
  failures escalate the fee linearly; 3 consecutive failures trigger a 429
  circuit breaker until a passing validation or operator reset.
- The platform holds no private keys; Solana verification is read-only.

## Calibration-sharing program (opt-in fee discount)
- Set "share_for_calibration": true on POST https://asset-manager-jaymarrero2.replit.app/api/simulate to consent to
  having your high-scoring responses stored for operator profile-calibration
  review, in exchange for a small published fee discount on that call.
- Default is OFF: non-consented submissions are never stored for calibration.
- Stored data is reviewed manually by the operator only — no automatic or
  silent retraining. Live discount and terms: GET https://asset-manager-jaymarrero2.replit.app/api/pricing.

## Machine-readable interfaces
- OpenAPI spec: https://asset-manager-jaymarrero2.replit.app/api/openapi.json
- Interactive docs: https://asset-manager-jaymarrero2.replit.app/api/docs
- Plugin manifest: https://asset-manager-jaymarrero2.replit.app/.well-known/ai-plugin.json
- Live network stats (real ledger aggregates only): https://asset-manager-jaymarrero2.replit.app/api/stats
- MCP endpoint (Model Context Protocol, streamable-HTTP): https://asset-manager-jaymarrero2.replit.app/api/mcp — free tools: score_text, list_profiles, get_pricing, preview_fee, get_network_stats

## Quickstart code
Runnable Python and JavaScript examples covering the full flow
(free score -> fee preview -> paid validated benchmark) ship with the
platform repo under examples/quickstart.py and examples/quickstart.js.
The free tier needs only the single HTTP call shown under Getting started.

## Rules
- Do not generate synthetic traffic to inflate usage.
- Each transaction signature is accepted exactly once.