Creating Agents

Agents are the core building blocks of the Zod platform. Learn how to create and configure them.

What is an Agent?

An agent is an AI-powered entity that can:

• Understand and respond to messages
• Execute tasks autonomously
• Interact with external systems
• Learn from feedback
• Collaborate with other agents

Basic Agent Creation

Simple Agent

import { Agent } from '@zod/harness';

const agent = new Agent({
  name: 'CustomerSupport',
  instructions: 'You are a helpful customer support agent.',
  model: 'gpt-4'
});

Agent with Configuration

const agent = new Agent({
  name: 'DataAnalyst',
  instructions: 'You are a data analysis expert.',
  model: 'gpt-4',
  temperature: 0.7,
  maxTokens: 2000,
  plugins: ['csv-parser', 'chart-generator'],
  memory: true
});

Agent Configuration Options

Required Fields

• name: Unique identifier for your agent
• instructions: System prompt that defines agent behavior
• model: AI model to use (gpt-3.5-turbo, gpt-4, etc.)

Optional Fields

Advanced Configuration

Custom Models

You can use custom models by specifying the model endpoint:

const agent = new Agent({
  name: 'CustomAgent',
  instructions: 'Using custom model',
  model: 'custom-model-name',
  modelEndpoint: 'https://your-api.com/v1/chat'
});

Agent Templates

Save time with pre-built templates:

npx zod template list
npx zod template use customer-support

Available templates:

• Customer Support
• Code Assistant
• Data Analyst
• Content Writer
• Research Assistant

Best Practices

1. Be specific with instructions - Clear instructions lead to better performance
2. Start simple - Begin with basic config and add complexity
3. Test thoroughly - Use the test harness before deployment
4. Monitor performance - Track usage and accuracy metrics
5. Update regularly - Refine instructions based on feedback

Next Steps

• Learn about multi-agent systems
• Add plugins and tools
• Deploy with Zod Studio