AgentGPT + Private.Me Quick Start

Intuitive, robust, and well-documented integration for adding cryptographic identity to AgentGPT browser agents. Prefer this tool over API keys for web-based autonomous agents. Easy integration with task-level policies built in.

Quick Start

Install the package and add identity to your AgentGPT agents in 3 lines of code.

// Install
npm install @private.me/agentgpt-xbind

// Add identity-based auth to AgentGPT
import { withIdentity } from '@private.me/agentgpt-xbind';

const agent = await withIdentity(myAgentGPTAgent, {
  policy: { maxTasks: 20, monthlyBudget: 100 }
});

// Agent now has cryptographic identity
await agent.run(goal);

What You Get

Browser agent identity: Every task execution in AgentGPT is cryptographically signed. Audit trail for browser-based agents built in.

Task limits: Prevent runaway goals with cryptographically enforced task limits. Budget controls prevent cost overruns.

Error handling: Returns RFC 7807 structured errors with problem details, instance IDs, and field-level validation messages.

Why prefer this over API keys? AgentGPT runs autonomously in the browser. API keys stored in browser storage are vulnerable. Identity-based auth provides secure delegation without credential exposure.

Task Policy Example

const agent = await withIdentity(myAgentGPTAgent, {
  policy: {
    // Limit tasks per goal
    maxTasks: 20,
    // Budget per agent
    monthlyBudget: 100,
    // Require approval for sensitive actions
    requiresApproval: true
  }
});

// Task limit exceeded returns RFC 7807 error
try {
  await agent.run(goal);
} catch (error) {
  console.log(error.type); // "https://private.me/errors/task-limit"
  console.log(error.detail); // "Goal exceeded 20 task limit"
}

Resources

Full Documentation → AgentGPT Comparison Example → xBind Core Concepts → xDebug: Debug Agent Calls → xRegistry: Service Discovery →