Documentation

MozAIc Documentation

Technical documentation, API references, and guides for using and integrating with MozAIc's multi-agent verification system.

Demo Only: This site is a demonstration of MozAIc's capabilities and is not a fully automated public system. The documentation describes the intended functionality of a future production system.

🚀

Getting Started

Quick start guide for new users

⚙️

API Reference

Complete API documentation

💡

Examples

Code samples and use cases

Getting Started

Overview

MozAIc is a multi-agent verification system that uses 6+ independent AI systems to analyze queries and anchor results on the Cardano blockchain. This guide will help you get started.

Key Concepts:

Query: A factual assertion you want to verify
Agent Response: Individual AI analysis with stance, confidence, and rationale
Consensus: Areas where multiple agents converge
Divergence: Areas where agents present different perspectives
Anchoring: Permanently storing verification hash on Cardano

Basic Workflow

Using the web interface is the simplest way to verify queries.

Submit a query — Enter your query in the text box (max 500 characters)
Select category — Choose from Science, Policy, Technology, or History
Add sources (optional) — Provide URLs for additional context
Enable privacy mode — Your query is hashed before processing
Pay and verify — Submit ~1.2 ADA on Cardano testnet
Wait for results — Typically takes 1-2 minutes
Review and export — Download JSON/CSV reports

API Reference

POST /v1/verify

Beta

Submit a query for verification

Request Body

{
  "query": "string (required, max 500 chars)",
  "category": "science | policy | technology | history",
  "sources": ["string[]"],
  "privacyMode": true
}

Response (202 Accepted)

{
  "jobId": "abc123...",
  "status": "queued",
  "estimatedWait": 120
}

Response (200 OK)

{
  "jobId": "abc123...",
  "status": "completed",
  "result": {
    "conclusion": "string",
    "confidence": 0.85,
    "agents": [...],
    "similarities": [...],
    "differences": [...],
    "txId": "abc...",
    "blockNumber": 12345678
  }
}

GET /v1/status/:jobId

Beta

Check verification status

Response

{
  "status": "queued | processing | completed | failed",
  "progress": 45,
  "currentStep": "Querying Agent 3 (Perplexity)",
  "result": null | {...}
}

GET /v1/results/:txId

Beta

Retrieve verification by transaction ID

Response

{
  "txId": "abc...",
  "blockNumber": 12345678,
  "timestamp": "2024-01-15T14:23:00Z",
  "query": "string",
  "result": {
    "conclusion": "string",
    "confidence": 0.85,
    "agents": [...],
    "similarities": [...],
    "differences": [...]
  }
}

System Architecture

Component Overview

1. Submission Gateway

Receives queries and validates input (length, format, rate limits). Creates job queue entries and generates unique job IDs for tracking.

Technologies: Node.js, Express, PostgreSQL

2. Agent Orchestrator

Manages parallel communication with 6+ AI agents (ChatGPT, Grok, Claude, Gemini, Perplexity, Deepseek). Handles retries, rate limits, and error recovery. Normalizes responses into consistent schema.

Technologies: Python, asyncio, LangChain

3. Consensus Analyzer

Analyzes agent responses to identify convergent themes and divergent perspectives using semantic similarity and NLP techniques. Generates synthesized conclusions.

Technologies: Python, scikit-learn, transformers

4. Blockchain Anchor

Creates compact cryptographic hashes of verification results and submits metadata to Cardano testnet using native assets. Tracks transaction confirmation.

Technologies: Cardano CLI, Haskell, Plutus

5. Report Generator

Formats results into human-readable reports with agent responses, consensus analysis, and blockchain verification. Exports to JSON and CSV formats.

Technologies: Node.js, Handlebars, csv-writer

Data Schema

Verification Result Schema

{
  "query": "string",
  "agents": [
    {
      "agent": "string",
      "analysis": "string",
      "confidence": 0.92,
      "references": ["string"],
      "sentiment": "supports | challenges | neutral"
    }
  ],
  "conclusion": "string",
  "confidence": 0.90,
  "similarities": [
    {
      "theme": "string",
      "agents": ["string"],
      "description": "string"
    }
  ],
  "differences": [
    {
      "theme": "string",
      "agents": ["string"],
      "description": "string"
    }
  ],
  "txId": "string",
  "blockNumber": 12345678
}

Examples

Basic Verification

const { verify } = require('@mozaic/sdk');

async function checkQuery() {
  const result = await verify({
    query: "Honey never spoils.",
    category: "science",
    privacyMode: true
  });
  
  console.log('Conclusion:', result.conclusion);
  console.log('Confidence:', result.confidence);
  console.log('TX ID:', result.txId);
}

checkQuery();

Frequently Asked Questions

How accurate are the verifications?

MozAIc doesn't assert absolute accuracy. Instead, it surfaces where multiple independent AI systems converge or diverge, providing a robust signal about consensus. The confidence score reflects the level of agreement among agents.

What happens if agents disagree?

Disagreement is valuable information. MozAIc highlights these areas in the "differences" section, showing where perspectives diverge and helping users understand uncertainty. This is often more useful than false certainty.

Is my query private?

With privacy mode enabled, only a cryptographic hash of your query is initially stored. The full text is revealed only after verification for reporting purposes. Future versions will support full ZK privacy via Midnight.

Can I add my own AI agent?

Yes! We're building an adapter framework for integrating custom AI agents. Agents must implement a standard interface for stance, confidence, rationale, and references. Contact us for details.

What's the cost?

Currently ~1.2 ADA on Cardano testnet, covering the blockchain transaction fee. This goes to network fees, not MozAIc. Future pricing on mainnet will be determined by governance.

Ready to Get Started?

Try the demo or read more about our implementation.

Try Demo Learn More