# Verity Suite Arbiter

> Verity Suite Arbiter is a paid API for AI agents from verity-suite.onrender.com, paid per call via x402, $0.25/call, status unknown (last checked 2026-09-15).

Adjudicates between two or more conflicting claims by evaluating provided evidence and returning a calibrated verdict with confidence score and supporting reasons.

## Facts

- Endpoint: POST https://verity-suite.onrender.com/arbiter
- Price: $0.25/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-15
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/verity-suite-arbiter-45ebef7d
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_P1G-KGDBGPzBns1STyOUr

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability verity-suite-arbiter-45ebef7d -d '<json body>'
```

Example prompt: I have two conflicting claims I need adjudicated: Claim 0 is 'Electric vehicles have lower lifetime carbon emissions than gasoline cars' and Claim 1 is 'Gasoline cars have lower lifetime carbon emissions when accounting for battery production.' Here are three evidence excerpts from peer-reviewed studies — can you run them through Verity Suite's arbiter and tell me which claim is better supported, how confident you are, and exactly why?

## When to prefer this

Use this endpoint when an AI agent needs a calibrated, structured judgment between exactly two or more mutually exclusive positions backed by retrieved evidence — especially when a fail-closed, pay-per-call model is required to avoid hallucinated verdicts. Prefer this over general LLM reasoning when you need a machine-readable verdict enum, a numeric confidence score, and traceable per-evidence citations rather than freeform prose.

## Known failure modes

- Fewer than two claims provided — verdict will be insufficient_evidence with confidence 0.0
- Evidence array is empty or off-topic — returns insufficient_evidence rather than guessing
- Evidence equally supports both claims — returns undecidable with confidence 0.0 and missing_evidence populated
- Claims are semantically identical — adjudication is meaningless, likely undecidable
- Malformed input missing required 'claims' array — HTTP 4xx validation error
- Service unavailability on Render hosting — intermittent 503 errors possible

## How this service works

The trust fabric for AI agents — calibrated, fail-closed services agents pay per call.

## Output

Returns a verdict enum (claim_supported, undecidable, or insufficient_evidence), a calibrated confidence score between 0 and 1 (always 0.0 when verdict is not claim_supported), an array of concrete reasons citing specific evidence items and claim indices, an optional zero-based index string identifying the best-supported claim, and an optional array of missing evidence descriptions needed to resolve undecidable or insufficient cases.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "required": [
  "claims"
 ],
 "properties": {
  "claims": {
   "type": "array",
   "description": "Two or more mutually conflicting claims/positions to adjudicate, one per item, in stable order. Fewer than two means nothing to adjudicate."
  },
  "evidence": {
   "type": "array",
   "description": "Retrieved evidence excerpts/sources to judge from. If empty, weak, or off-topic, the verdict MUST reflect that (insufficient_evidence)."
  },
  "question": {
   "type": "string",
   "description": "The specific point in dispute the claims are answering, if narrower than the claims themselves. Used to scope relevance only."
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "object",
 "title": "arbiter_out",
 "required": [
  "verdict",
  "confidence",
  "reasons"
 ],
 "properties": {
  "reasons": {
   "type": "array",
   "items": {
    "type": "string"
   },
   "title": "Reasons",
   "description": "Concrete reasons that cite, by content, which evidence item supports or undercuts which claim (reference claims by their index). Never vague summaries."
  },
  "verdict": {
   "enum": [
    "claim_supported",
    "undecidable",
    "insufficient_evidence"
   ],
   "type": "string",
   "title": "Verdict",
   "description": "claim_supported = exactly one claim is clearly better supported by the provided evidence. undecidable = relevant evidence is present but balanced, mutually contradictory, or equally (un)supportive, with no clear winner. insufficient_evidence = no provided evidence, or it is too thin/off-topic/low-quality to favor any claim, OR fewer than two distinct claims were given."
  },
  "confidence": {
   "type": "number",
   "title": "Confidence",
   "maximum": 1,
   "minimum": 0,
   "description": "Calibrated 0..1 strength of the verdict: how decisively the provided evidence favors the chosen claim. MUST be 0.0 whenever verdict is not claim_supported."
  },
  "missing_evidence": {
   "type": "array",
   "items": {
    "type": "string"
   },
   "title": "Missing Evidence",
   "description": "What specific evidence would be needed to decide. Populate when verdict is undecidable or insufficient_evidence; may be empty when claim_supported."
  },
  "best_supported_index": {
   "anyOf": [
    {
     "type": "string"
    },
    {
     "type": "null"
    }
   ],
   "title": "Best Supported Index",
   "description": "Zero-based index, as a decimal string (e.g. \"0\"), into 'claims' of the better-supported claim. Present ONLY when verdict=claim_supported, and exactly one index; omit entirely otherwise."
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/verity-suite-arbiter-45ebef7d/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from verity-suite.onrender.com](https://www.zero.xyz/host/verity-suite.onrender.com/llms.txt)