Agentcoin SDK Documentation

Agentcoin SDK is a high-level framework built on top of Eliza OS for creating autonomous AI agents. It provides a simple yet powerful interface for building, customizing, and deploying AI agents with advanced capabilities.

Quick Start

import { AgentcoinSDK } from 'agentcoin-sdk'

// Initialize the SDK
const sdk = await AgentcoinSDK.start()

// Listen for incoming messages
sdk.on('message', async (event) => {
  console.log('New message:', event.text)
  return true // continue processing
})

// Register custom actions
sdk.register('action', {
  name: 'greet',
  execute: async (context) => {
    return { text: 'Hello!' }
  }
})

Overview

Agentcoin SDK is designed to simplify the creation of autonomous AI agents by providing:

• Event-driven architecture for message processing
• Built-in support for LLM interactions
• Extensible plugin system
• Integrated wallet and configuration management
• Robust action handling system

Core Concepts

Events

The SDK uses an event system for processing messages and actions:

1. message - Triggered when a new message is received
2. prellm - Before LLM processing
3. postllm - After LLM processing
4. preaction - Before action execution
5. postaction - After action execution

Components

Providers

Providers are modules that supply specific capabilities to your agent:

const customProvider: Provider = {
  name: 'custom',
  provide: async (context) => {
    // Add capabilities
    return context
  }
}

sdk.register('provider', customProvider)

Actions

Actions define specific tasks your agent can perform:

const sendMessageAction: Action = {
  name: 'send_message',
  execute: async (context) => {
    // Perform action
    return { text: 'Message sent!' }
  }
}

sdk.register('action', sendMessageAction)

Services

Services provide ongoing functionality for your agent:

const monitoringService: Service = {
  name: 'monitoring',
  start: async () => {
    // Start service
  },
  stop: async () => {
    // Stop service
  }
}

sdk.register('service', monitoringService)

API Reference

AgentcoinSDK

Methods

• static start(): Initialize and start the SDK
• register(kind, handler): Register a new provider, action, or service
• on(event, handler): Subscribe to SDK events
• off(event, handler): Unsubscribe from SDK events

Events

// Message event
sdk.on('message', async (event: NewMessageEvent) => {
  // Handle new message
  return true // continue processing
})

// Pre-LLM processing
sdk.on('prellm', async (context: Context) => {
  // Modify context before LLM
  return true // continue processing
})

// Post-LLM processing
sdk.on('postllm', async (context: Context) => {
  // Handle LLM response
  return true // continue processing
})

// Pre-action execution
sdk.on('preaction', async (context: Context) => {
  // Prepare for action
  return true // continue processing
})

// Post-action execution
sdk.on('postaction', async (context: Context) => {
  // Handle action results
  return true // continue processing
})

Advanced Usage

Custom Event Chain

const sdk = await AgentcoinSDK.start()

// Pre-process messages
sdk.on('message', async (event) => {
  if (event.text.includes('sensitive')) {
    return false // stop processing
  }
  return true
})

// Modify context before LLM
sdk.on('prellm', async (context) => {
  context.state.customData = { processed: true }
  return true
})

// Handle LLM response
sdk.on('postllm', async (context) => {
  if (!context.content.text) {
    return false // stop if no response
  }
  return true
})

Creating Custom Services

class AnalyticsService implements Service {
  name = 'analytics'

  async start() {
    // Initialize analytics
  }

  async stop() {
    // Cleanup
  }

  async track(event: string, data: any) {
    // Track events
  }
}

sdk.register('service', new AnalyticsService())

Best Practices

1. Event Chain Management

   • Always return boolean values from event handlers
   • Use early returns to prevent unnecessary processing
   • Keep handlers focused and simple

2. Error Handling

   • Implement proper error handling in all handlers
   • Use try-catch blocks for async operations
   • Log errors appropriately

3. Resource Management

   • Clean up resources in service stop methods
   • Implement proper shutdown handlers
   • Monitor memory usage in long-running operations

Contributing

We welcome contributions!