Creating an Agent

This guide walks you through creating an Agent RDA for workflow automation.

What is an Agent RDA?

An Agent RDA orchestrates multi-step workflows by communicating with external services via webhooks.

Prerequisites

xpay✦ Hub account
Webhook endpoint (n8n, Make, custom server)
Understanding of HTTP APIs

Step 1: Set Up Your Webhook

Using n8n

Create a new workflow in n8n
Add a Webhook trigger node
Configure to receive POST requests
Copy the webhook URL

Webhook Payload

Your endpoint will receive:


{
  "run_id": "run_abc123",
  "rda_id": "rda_xyz",
  "rda_slug": "my-agent",
  "rda_name": "My Agent",
  "inputs": {
    "field1": "value1",
    "field2": "value2"
  },
  "user_id": "did:privy:xxx",
  "timestamp": 1702000000000
}

Response Format

Return results as JSON:


{
  "status": "success",
  "output": "Your result here",
  "metadata": {
    "steps_completed": 3,
    "duration_ms": 5000
  }
}

Step 2: Define Your RDA

Basic Information

Name - Clear, action-oriented title
Slug - URL-friendly identifier
Description - What the agent does
Tags - Relevant keywords for discovery

Step 3: Configure Agent Settings


const agentConfig = {
  webhookUrl: 'https://your-n8n.com/webhook/xxx',
  webhookMethod: 'POST',
  webhookHeaders: {
    'X-Custom-Header': 'value'
  },
  timeoutSeconds: 60,
  maxRetries: 3,
  retryDelayMs: 1000,
  // For async/long-running tasks
  pollUrl: 'https://your-n8n.com/poll/xxx',
  pollIntervalMs: 2000
}

Configuration Options

Option	Description	Default
`webhookUrl`	Endpoint to call	Required
`webhookMethod`	HTTP method	POST
`webhookHeaders`	Custom headers
`timeoutSeconds`	Max execution time	60
`maxRetries`	Retry attempts	3
`pollUrl`	Async polling endpoint	null
`pollIntervalMs`	Polling interval	2000

Step 4: Design Input Schema

Define inputs users will provide:


const inputSchema = [
  {
    name: 'targetCompany',
    label: 'Company Name',
    type: 'text',
    required: true,
    placeholder: 'Acme Inc'
  },
  {
    name: 'action',
    label: 'Action',
    type: 'select',
    required: true,
    options: ['research', 'analyze', 'report']
  },
  {
    name: 'depth',
    label: 'Research Depth',
    type: 'number',
    required: false,
    default: 3
  }
]

Step 5: Handle Async Workflows

For long-running tasks, use polling:

Initial Response


{
  "status": "pending",
  "job_id": "job_123",
  "message": "Processing started"
}

Poll Endpoint

The poll endpoint is called until completion:


{
  "status": "completed",
  "output": "Final results here",
  "metadata": {...}
}

Status Values

pending - Still processing
completed - Success, results ready
failed - Error occurred

Step 6: Set Pricing

Agents use custom pricing:


const pricing = {
  model: 'per-run',
  amount: 0.10, // $0.10 per run
  currency: 'USDC'
}

Consider:

Your infrastructure costs
External API costs
Execution time
Value delivered

Step 7: Testing

Before publishing:

Test the webhook - Ensure it handles all inputs
Test timeouts - Verify behavior on slow responses
Test errors - Handle failures gracefully
Test polling - For async workflows

Example: Lead Research Agent


// RDA Configuration
{
  name: 'Lead Research Agent',
  slug: 'lead-research',
  type: 'agent',
 
  agentConfig: {
    webhookUrl: 'https://n8n.example.com/webhook/lead-research',
    webhookMethod: 'POST',
    timeoutSeconds: 120,
    pollUrl: 'https://n8n.example.com/poll/lead-research',
    pollIntervalMs: 5000
  },
 
  inputSchema: [
    {
      name: 'companyName',
      label: 'Company to Research',
      type: 'text',
      required: true
    },
    {
      name: 'researchDepth',
      label: 'Research Depth',
      type: 'select',
      options: ['quick', 'standard', 'deep'],
      default: 'standard'
    }
  ],
 
  pricing: {
    model: 'per-run',
    amount: 0.15,
    currency: 'USDC'
  }
}