AgentGPT + xLink Quickstart — PRIVATE.ME Platform

Step 01

Installation

Install the AgentGPT + xLink integration package via npm, pnpm, or yarn.

TERMINAL

# npm
npm install @private.me/agentgpt-xlink

# pnpm
pnpm add @private.me/agentgpt-xlink

# yarn
yarn add @private.me/agentgpt-xlink

ZERO DEPENDENCIES

This package has zero npm runtime dependencies. All cryptography uses the Web Crypto API. Works in Node.js, Deno, Bun, Cloudflare Workers, and browser environments.

Step 02

Basic Example

Replace OAuth tokens with persistent identity in your AgentGPT browser automation workflows.

Before: Traditional OAuth

TYPESCRIPT — BEFORE (OAuth Tokens)

import { Agent } from 'agentgpt';

// OAuth flow: refresh token → access token → API call
const agent = new Agent({
  auth: {
    clientId: process.env.OAUTH_CLIENT_ID,
    clientSecret: process.env.OAUTH_CLIENT_SECRET,
    refreshToken: process.env.REFRESH_TOKEN
  }
});

// Every request requires token refresh check
await agent.execute('Navigate to dashboard');
// ⚠️ If token expires mid-workflow, entire agent restarts

After: xLink Identity

TYPESCRIPT — AFTER (xLink Identity)

import { Agent } from 'agentgpt';
import { withXLink } from '@private.me/agentgpt-xlink';

// DID identity generated once, valid forever
const agent = await withXLink(new Agent(), {
  agentName: 'browser-automation-01'
});

// No token refresh, no expiry, no cascades
await agent.execute('Navigate to dashboard');
// ✅ Identity persists across restarts

PERSISTENT IDENTITY

xLink generates a DID identity (Ed25519 keypair) for each agent. The identity is stored locally and reused across sessions. No OAuth tokens, no expiry, no refresh.

Full Browser Automation Example

TYPESCRIPT — COMPLETE WORKFLOW

import { Agent } from 'agentgpt';
import { withXLink, createTrustRegistry } from '@private.me/agentgpt-xlink';

// Step 1: Create trust registry (shared across all agents)
const registry = await createTrustRegistry();

// Step 2: Create browser automation agent with identity
const agent = await withXLink(new Agent(), {
  agentName: 'browser-automation-01',
  trustRegistry: registry
});

// Step 3: Execute tasks — identity used automatically
await agent.execute('Open Google Sheets');
await agent.execute('Navigate to Sales Dashboard');
await agent.execute('Export data to CSV');

// ✅ No token refresh, no OAuth flow, no cascades
// ✅ Agent identity persists across browser restarts
// ✅ 91ms authentication (vs 54,853ms OAuth flow)

Step 03

Why xLink for AgentGPT

Three core problems solved: cascading failures, slow authentication, and token management overhead.

Problem 1: Cascading Failures

One expired OAuth token can restart 500+ browser agents simultaneously. The refresh token expires → all access tokens invalidate → every running agent crashes → 500 agents attempt to refresh at once → API rate limit triggers → cascading failure propagates across your entire automation fleet.

REAL INCIDENT

2:47 AM: Refresh token expires for shared service account.
2:48 AM: 500 browser agents detect invalid access tokens, attempt simultaneous refresh.
2:49 AM: OAuth provider rate limit triggers (429 Too Many Requests).
2:50 AM: All 500 agents crash. Manual intervention required.

xLink eliminates cascades. Every agent has a unique DID identity with no expiry. There is no shared OAuth token, so there is no single point of failure. One agent failure never propagates to others.

Problem 2: Slow Authentication

54,853ms

OAuth Flow

91ms

xLink Identity

603×

Faster

OAuth token refresh requires: check access token expiry → HTTP request to OAuth provider → wait for new token → retry original request. xLink identity verification is a local Ed25519 signature check. No network round trip. 603× faster.

Problem 3: Token Management Overhead

Task	OAuth Tokens	xLink Identity
Store credentials	Encrypted vault + rotation	Local Ed25519 keypair
Refresh flow	Every 60 minutes	Never (no expiry)
Revocation	Manual via OAuth provider UI	Remove from trust registry
Multi-agent setup	Share refresh token OR create 500 OAuth apps	Each agent gets unique DID

Step 04

Next Steps

You now have persistent identity for AgentGPT browser agents. Here's what to explore next.

1. Trust Registry Management

Control which agents can authenticate to your services. The trust registry is a simple allowlist of DID identities. Add new agents, revoke compromised ones, and enforce capability-based access control.

TYPESCRIPT — TRUST REGISTRY

import { createTrustRegistry } from '@private.me/agentgpt-xlink';

const registry = await createTrustRegistry();

// Add new agent to allowlist
await registry.add('did:key:z6MkpTHz...', {
  capabilities: ['browser:navigate', 'sheets:export']
});

// Revoke compromised agent
await registry.revoke('did:key:z6MkpTHz...');

2. Multi-Agent Orchestration

Scale to hundreds of browser agents with independent identities. No shared OAuth tokens, no cascading failures, no rate limit collisions.

TYPESCRIPT — 500 AGENTS

const agents = await Promise.all(
  Array.from({ length: 500 }, async (_, i) => {
    return await withXLink(new Agent(), {
      agentName: `browser-agent-${i}`,
      trustRegistry: registry
    });
  })
);

// All 500 agents execute in parallel — no token conflicts
await Promise.all(agents.map(a => a.execute('Run daily report')));

3. Read the Full White Paper

Learn about the underlying cryptographic protocol, benchmark methodology, and production deployment patterns.

xLink Technical White Paper →

4. Explore Other ACIs

Private.Me offers 224 ACIs (Agent-Connectable Interfaces) for different use cases. All use the same identity protocol.

Browse All ACIs →