False-consensus check across 2 to 16 agent answers
consensusFalse-consensus check for 2 to 16 answers to the same task, so a group of agents does not approve a spend just because they copied each other. It scores how the answers agree and flags agreement that looks copied or too thin to trust. Screwtape is not tied to the agent framework or the model that wrote the answers. Call it when several agents, samples or tools answered the same question and you are about to accept the majority. Do NOT call it with a single answer (use judge), with answers to different tasks, or as proof that an answer is correct: high agreement is not evidence. Inputs: task, answers (2 to 16 objects {agent_id?: string, output: string}), optional evidence (a string or an array of strings, used when scoring), retain, and optional call-level agent_id and version tags. The call-level agent_id is not the per-answer agent_id. Returns: consensus_answer, agreement_score, false_consensus_risk (low, medium or high), reasons, clusters, engine, request_id, retention, the tags when sent, and receipt when signing is configured. In that receipt, verdict is false_consensus_risk and score is agreement_score. If the risk is medium or high, verify consensus_answer with judge before using it. next_step.action is pass for low risk and escalate otherwise. Cost: 1 call. First 3 calls per UTC day need no key; then use Authorization: Bearer or pay per call. A free key covers 10 calls per UTC day (get_api_key or https://screwtape.ai/signup/). After the shared 3/day anonymous allowance (3/minute, global 2000/day), tools/call without a key or payment is HTTP 200 with isError true and the payment requirements in structuredContent.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"task": "<task>",
"answers": "<answers>"
},
"name": "consensus"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| mode | No | Optional. blocking (the default) is today's behaviour. advisory computes the same verdict, checks, next_step and receipt, then adds mode, decision_owner human, recommended_action (approve, review or return) and a notice. The receipt signs mode advisory as schema_version 5. Same price. | |
| task | Yes | The question all answers respond to. | |
| share | No | Optional. false keeps this call out of opt-in data sharing even when the key has sharing on. It cannot turn sharing on. | |
| retain | No | Optional. true stores this call for 24 hours; requires an API key. Default false. | |
| answers | Yes | 2 to 16 answers. Each is {agent_id (optional, string), output (string)}. Use one entry per agent or sample. | |
| version | No | Optional version tag for that agent. At most 64 characters. | |
| agent_id | No | Optional tag for the agent making this call. 1 to 64 characters: letters, digits, and . _ : -. This is not the per-answer agent_id on a consensus answer. | |
| evidence | No | Optional passages the agreed answer should be grounded in. A string is one passage. An array holds up to 20 passages of 4000 characters. Used when scoring false-consensus risk. |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| mode | No | Present only when the call sent mode: advisory. | |
| model | No | ||
| engine | No | groq, workers_ai, deterministic, or demo. | |
| notice | No | Advisory only. Fixed text: the call does not approve, hold or release any payment. | |
| reasons | Yes | ||
| receipt | No | ||
| version | No | Echoed when the call sent a version tag. | |
| agent_id | No | Echoed when the call sent an agent_id tag. Not the per-answer agent_id. | |
| clusters | No | ||
| degraded | No | ||
| next_step | No | ||
| retention | No | ||
| request_id | Yes | ||
| verification | No | ||
| decision_owner | No | Advisory only. A person owns the decision. | |
| receipt_reason | No | ||
| agreement_score | Yes | null when no model scored agreement (degraded). | |
| consensus_answer | No | ||
| recommended_action | No | Advisory only, deterministic and not signed. Judge: next_step pass is approve, escalate is review, retry is return (send back to the agent). Consensus: risk low is approve, medium or high is review. | |
| false_consensus_risk | Yes | ||
| unsupported_agreement | No |