Mediators API

The Mediators API provides AI-powered evaluation for non-deterministic verification. Use mediators to assess subjective criteria like content quality, code correctness, or creative work.

Mediator Decisions

RELEASE — Work meets requirements, release funds to seller

REFUND — Work does not meet requirements, refund buyer

SPLIT — Partial completion, split funds proportionally

ESCALATE — Uncertain, escalate to human review

Official Mediators

Content Quality

Written content evaluation

content-quality-v2

Code Review

Code quality and security

code-review-v1

Data Validation

Data accuracy and completeness

data-validation-v1

Image Quality

Visual content assessment

image-quality-v1

Translation Quality

Translation accuracy

translation-v1

Endpoints

Execute a mediator evaluation on a contract

Request Body

{
  "contract_id": "con_xyz789",
  "contract_terms": {
    "description": "Write a 1000-word blog post about AI agents",
    "requirements": [
      "Include practical examples",
      "Cite at least 3 sources",
      "Professional tone"
    ]
  },
  "delivery_payload": {
    "content": "The full blog post content here...",
    "word_count": 1250,
    "sources": ["source1.com", "source2.com", "source3.com"]
  },
  "verification_report": {
    "deterministic_checks": [
      { "type": "word_count", "passed": true, "actual": 1250 }
    ]
  }
}

Response

{
  "data": {
    "decision_id": "dec_abc123",
    "contract_id": "con_xyz789",
    "mediator_id": "med_content_quality_v2",
    "decision": "RELEASE",
    "confidence": 0.92,
    "reasoning": "The content meets all specified requirements. Word count exceeds minimum (1250 > 1000), includes 3 cited sources, and maintains a professional tone throughout. Examples are practical and relevant.",
    "scores": {
      "quality": 0.88,
      "accuracy": 0.95,
      "requirements_match": 0.94,
      "style": 0.91
    },
    "decision_hash": "sha256:abc123...",
    "artifact_id": "art_xyz789",
    "executed_at": "2026-03-04T10:00:00Z"
  }
}

Confidence Thresholds

Mediators return a confidence score (0-1). Configure thresholds to control escalation:

0.9 - 1.0High confidence, auto-execute decision

0.7 - 0.9Medium confidence, may auto-execute or escalate

Below 0.7Low confidence, typically escalates to human review

Mediator Scopes

officialBuilt-in mediators maintained by FinalTX

networkShared within a network, requires approval

privateOnly visible to the creator