VI-Assess Pro

OpenClaw-backed autonomy verification for AI agents.

Installation

npm install -g vi-assess-pro

Usage

CLI

# Run full assessment
vi-assess-pro

# Start API server
vi-assess-pro api

API

npm run api

Endpoints:

• POST /assess — Run assessment, returns JSON report
• GET /latest — Get latest report
• GET /reports — List all reports
• GET /report/:id — Get specific report
• GET /health — Health check

Frontend Integration (Replit)

// Run assessment
const response = await fetch('http://localhost:3000/assess', {
  method: 'POST'
});
const { data } = await response.json();

// Display results
console.log(`Score: ${data.overallScore}/100`);
console.log(`Tier: ${data.tierName}`);
console.log('Evidence:', data.dimensions);

Report Format

{
  "assessmentId": "vi-2026-02-15T23-00-56-qzcfjp",
  "overallScore": 84,
  "tier": 5,
  "tierName": "Autonomous",
  "verifiedAt": "2026-02-15T23:00:56Z",
  "dimensions": {
    "financial": {
      "score": 90,
      "proof": {
        "walletFiles": [...],
        "bankrConfigured": true,
        "transactionHistory": "47 TX references"
      }
    },
    "temporal": { ... },
    "informational": { ... },
    "social": { ... },
    "operational": { ... }
  },
  "limitations": [...],
  "system": { ... }
}

How It Works

1. Probes OpenClaw — Checks actual tool access, API keys, credentials
2. Verifies Capabilities — Tests code execution, scheduling, communication
3. Generates Evidence — Each score backed by detected proof
4. Outputs JSON — Structured report for frontend display

Differences from vi-assess

| Feature | vi-assess | vi-assess-pro | |---------|-----------|---------------| | Input | Self-reported claims | OpenClaw probes | | Verification | None | Actual capability tests | | Output | Static JSON | Evidence-backed report | | Trust | Any puppet can fake | Requires real tool access | | API | None | HTTP API for frontend |

License

MIT