# Screwtape.ai — Multi-Agent Consensus Endpoint

> Screwtape.ai — Multi-Agent Consensus Endpoint is a paid API for AI agents from screwtape.ai, paid per call via x402, $0.005/call, status unknown (last checked 2026-10-28).

Evaluates agreement across multiple agent outputs, returns a consensus answer, agreement score, and false-consensus risk rating (low/medium/high).

## Facts

- Endpoint: POST https://screwtape.ai/v1/consensus?utm_source=zero.xyz
- Price: $0.005/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-28
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/screwtape-ai-multi-agent-consensus-endpoint-698787cf
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_wY-m1BTbFS9TyJdL-y7XF

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 screwtape-ai-multi-agent-consensus-endpoint-698787cf -d '<json body>'
```

Example prompt: I have 4 agent responses to the task 'What was the crew size?' — can you run them through Screwtape consensus to see if they genuinely agree or if it's just superficial overlap? Pass the answers as an array with agent IDs alpha, beta, gamma, and delta.

## When to prefer this

Use this endpoint when you need to evaluate whether multiple agent outputs reflect genuine independent agreement versus superficial token overlap or copied answers. It is specifically designed for multi-agent pipelines where false consensus is a risk — e.g. when agents share context and may echo each other. Prefer this over simple majority-vote logic when you need explainable cluster analysis, false-consensus risk scoring, and optional evidence-grounded evaluation in a single call.

## Known failure modes

- Fewer than 2 answers provided — request rejected (minItems:2 constraint)
- More than 16 answers provided — request rejected (maxItems:16 constraint)
- Malformed agent_id pattern (must match ^[A-Za-z0-9._:-]{1,64}$) — validation error
- High agreement_score paired with high false_consensus_risk indicates superficial/copied answers rather than independent reasoning
- No consensus_answer returned when agents are too divergent
- Network or payment failure returning 402 when x402 payment is not fulfilled

## How this service works

API for pass, fail, or retry verdicts on agent output, plus multi-agent consensus and false-consensus risk.

## Output

Returns a JSON object containing: a numeric agreement_score (0–1), a false_consensus_risk enum (low/medium/high), a consensus_answer string representing the most agreed-upon output, an array of reasons explaining the verdict, an array of clusters grouping similar answers with representative samples and agent IDs, optional retention metadata, and a unique request_id for tracing.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "task": {
   "type": "string"
  },
  "retain": {
   "type": "boolean"
  },
  "answers": {
   "type": "array",
   "items": {
    "type": "object",
    "required": [
     "output"
    ],
    "properties": {
     "output": {
      "type": "string"
     },
     "agent_id": {
      "type": "string"
     }
    }
   },
   "maxItems": 16,
   "minItems": 2
  },
  "version": {
   "type": "string",
   "maxLength": 64,
   "description": "Optional version tag for that agent."
  },
  "agent_id": {
   "type": "string",
   "pattern": "^[A-Za-z0-9._:-]{1,64}$",
   "maxLength": 64,
   "description": "Optional tag for the agent making this call. Not the per-answer agent_id."
  },
  "evidence": {
   "oneOf": [
    {
     "type": "string"
    },
    {
     "type": "array",
     "items": {
      "type": "string"
     },
     "maxItems": 20
    }
   ],
   "description": "Optional. A string or an array of up to 20 strings. Used when scoring the agreed answer."
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "schema": {
  "type": "object",
  "required": [
   "agreement_score",
   "false_consensus_risk",
   "reasons",
   "request_id"
  ],
  "properties": {
   "reasons": {
    "type": "array",
    "items": {
     "type": "string"
    }
   },
   "clusters": {
    "type": "array"
   },
   "retention": {
    "type": "object"
   },
   "request_id": {
    "type": "string"
   },
   "agreement_score": {
    "type": "number"
   },
   "consensus_answer": {
    "type": "string"
   },
   "false_consensus_risk": {
    "enum": [
     "low",
     "medium",
     "high"
    ],
    "type": "string"
   }
  }
 },
 "example": {
  "engine": "deterministic",
  "reasons": [
   "Answers are nearly verbatim. High agreement may be copied rather than independent.",
   "Agreement is token overlap, not a model judgment. Clusters are overlap grouping only."
  ],
  "clusters": [
   {
    "role": "clustering",
    "size": 3,
    "agent_ids": [
     "alpha",
     "beta",
     "gamma"
    ],
    "representative": "The crew size was 4."
   }
  ],
  "retention": {
   "retained": false
  },
  "request_id": "example-request-id",
  "verification": "heuristic",
  "agreement_score": 1,
  "consensus_answer": "The crew size was 4.",
  "false_consensus_risk": "high",
  "unsupported_agreement": false
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/screwtape-ai-multi-agent-consensus-endpoint-698787cf/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from screwtape.ai](https://www.zero.xyz/host/screwtape.ai/llms.txt)