Decision Engine API

Get a decision, not an answer.

Resolve uncertainty via structured multi-model disagreement. 5 frontier AIs debate your question and return a verdict with confidence score, ranked recommendations, and cost transparency.

Quick Start

One POST request. Structured JSON response. That's it.

curl -X POST https://api.pronnpt.com/v1/debate/run \
  -H "Authorization: Bearer prn_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Should my startup pivot to AI?",
    "speed": "fast"
  }'

Base URL

https://api.pronnpt.com/v1

All API endpoints are versioned under /v1.

Endpoints

Parameters

POST body for /v1/debate/run

Speed Modes

Set speed directly, or set max_latency_ms and we'll pick the fastest mode that fits your budget.

Available Models

OpenRouter prices per 1M tokens. We charge exactly what OpenRouter charges us — zero markup.

Use Cases

Trading decisions, product strategy, legal reasoning, clinical judgement, research synthesis, policy analysis.

Code Examples

curl -X POST https://api.pronnpt.com/v1/debate/run \
  -H "Authorization: Bearer prn_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Should my startup pivot to AI?",
    "speed": "fast"
  }'

Summary Mode

Set response_mode: "summary" to get just the decision, confidence, and cost. Perfect for agents that only need the verdict.

// Minimal response — just the decision
const res = await fetch("https://api.pronnpt.com/v1/debate/run", {
  method: "POST",
  headers: {
    "Authorization": "Bearer prn_your_key_here",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    question: "Should we acquire this company?",
    speed: "fast",
    response_mode: "summary"
  })
});

// Returns:
// {
//   "debate_id": 305,
//   "type": "proposition",
//   "decision": "REJECT",
//   "confidence": 0.60,
//   "executive_summary": "The council advises...",
//   "cost_usd": 0.02,
//   "tokens_used": 3100
// }

Response Format

Proposition debates return a verdict with decision (ACCEPT/REJECT/SPLIT) and numeric confidence. Strategy debates return a synthesis with ranked recommendations.

{
  "debate_id": 301,
  "type": "proposition",
  "status": "complete",
  "models": ["claude", "gpt", "grok"],
  "speed": "fast",
  "verdict": {
    "decision": "ACCEPT",
    "confidence": 0.85,
    "vote_split": { "for": 2, "against": 1 },
    "executive_summary": "The council recommends..."
  },
  "chair_summary": {
    "recommendation": "...",
    "consensus": "...",
    "dissent": "..."
  },
  "follow_up_questions": ["What about...", "Have you considered..."],
  "tokens_used": 4200,
  "cost_usd": 0.03,
  "report_url": "https://pronnpt.com/debate/301"
}

{
  "debate_id": 302,
  "type": "strategy",
  "status": "complete",
  "synthesis": {
    "executive_summary": "The council recommends...",
    "ranked_recommendations": [
      {
        "rank": 1,
        "title": "Use PostgreSQL with TimescaleDB",
        "description": "Combine PostgreSQL reliability with...",
        "proposed_by": "Claude",
        "total_score": 5,
        "avg_rank": 1.25
      }
    ],
    "confidence": 0.85
  },
  "tokens_used": 5100,
  "cost_usd": 0.04,
  "report_url": "https://pronnpt.com/debate/302"
}

Pricing

You pay exactly what OpenRouter charges us for tokens. Zero markup. We may introduce a small margin in the future, but today it's pure pass-through.

API Keys

Authentication

Pass your API key in the Authorization header:

Authorization: Bearer prn_your_key_here

Without a key: 50 free requests per IP. With a key: starter tier at 100/day, scales on request.

Rate Limits

Questions? ben@pronnpt.com