@agentgatepkg/client

Browser SDK for AgentGate — AI agent verification with fingerprinting, telemetry, and challenge solving.

Install

npm install @agentgatepkg/client

Quick Start

import { AgentGateClient } from "@agentgatepkg/client";

const client = new AgentGateClient({
  endpoint: "http://localhost:8080",
});

const result = await client.verify();

console.log(result.isAgent);    // true
console.log(result.agentScore); // 0.94
console.log(result.token);      // JWT for authenticated requests

How It Works

1. Fingerprinting — Collects browser fingerprint (canvas, WebGL, fonts, navigator)
2. Telemetry — Records mouse, keyboard, and scroll behavior patterns
3. Challenge — Requests and solves a machine-oriented challenge from the server
4. Scoring — Submits all data for multi-factor agent/human scoring
5. Verdict — Returns verification result with JWT token if verified

Configuration

const client = new AgentGateClient({
  endpoint: "https://your-agentgate-server.com", // Required
  apiKey: "ag_xxxxx",          // Optional: API key for auth
  autoVerify: true,            // Auto-verify on init (default: true)
  telemetry: true,             // Collect behavioral telemetry (default: true)
  fingerprint: true,           // Collect browser fingerprint (default: true)
  onVerified: (result) => {},  // Callback on successful verification
  onRejected: (result) => {},  // Callback on rejection
  onError: (error) => {},      // Callback on error
});

await client.init();    // Start collection + auto-verify
client.destroy();       // Cleanup

API

| Method | Description | |--------|-------------| | init() | Initialize client, start telemetry, auto-verify | | verify() | Run the full verification flow manually | | destroy() | Stop telemetry and cleanup resources |

Requirements

• AgentGate backend running (setup guide)

Links

• GitHub
• Full Documentation

License

MIT