Skip to Content
Functions.do is released 🎉
Overview

Agents.do  - Autonomous Digital Workers

Intelligent autonomous workers that accomplish business tasks with elegance

Overview

Agents.do is a core primitive of the .do  ecosystem, providing a clean, type-safe interface for creating and managing autonomous digital workers. These intelligent agents combine the power of functions, workflows, and integrations to accomplish complex business tasks with minimal human intervention.

The Challenge

Building autonomous business systems presents several challenges:

  • Decision Complexity: Enabling AI to make appropriate business decisions
  • Tool Orchestration: Coordinating multiple capabilities for complex tasks
  • Context Awareness: Maintaining relevant context across interactions
  • Human Collaboration: Seamlessly working alongside human operators
  • Adaptability: Responding to changing business conditions

Key Features

  • Elegant Agent Design - Clean, intuitive interfaces for autonomous workers
  • Seamless Tool Integration - Effortless access to functions, workflows, and integrations
  • Intelligent Decision-Making - Autonomous operation within defined parameters
  • Minimal Configuration - Simple setup with sensible defaults
  • Human Collaboration - Graceful handoffs between AI and human operators

Elegant API Design

Agents.do provides a clean, intuitive interface for creating and managing autonomous digital workers:

import { Agent } from 'agents.do'
 
// Simple initialization with default settings
const agent = new Agent({
  name: 'customer-support',
  description: 'Helps customers with product questions and technical issues',
})
 
// Or with custom configuration
const agent = new Agent({
  name: 'research-assistant',
  description: 'Conducts in-depth research on specified topics',
  model: 'openai/gpt-4o',
  memory: {
    type: 'conversational',
    retention: '30d',
  },
})

Tool Integration

A powerful capability of Agents.do is the elegant integration with other .do  services as tools:

import { Agent } from 'agents.do'
import { AI } from 'functions.do'
 
// Define AI functions
const ai = AI({
  generateSummary: {
    text: 'string',
    maxLength: 'number',
    summary: 'string',
  },
 
  analyzeData: {
    data: 'any[]',
    metrics: 'string[]',
    results: 'Record<string, number>',
  },
})
 
// Create an agent with access to functions as tools
const researchAgent = new Agent({
  name: 'research-assistant',
  description: 'Conducts comprehensive research on business topics',
  tools: [ai.generateSummary, ai.analyzeData, ai.searchWeb, ai.requestHumanVerification],
})
 
// Create an agent with access to workflows as tools
const customerSupportAgent = new Agent({
  name: 'support-assistant',
  description: 'Helps customers with product issues and service requests',
  tools: [workflows.escalationProcess, workflows.refundRequest, workflows.knowledgeBaseUpdate],
})
 
// Create an agent with access to searches as tools
const knowledgeAgent = new Agent({
  name: 'knowledge-assistant',
  description: 'Provides accurate information from company knowledge bases',
  tools: [searches.productCatalog, searches.customerRecords, searches.knowledgeBase],
})

Intelligent Interactions

Agents.do enables elegant, context-aware interactions with users and systems:

// Handle a customer inquiry with minimal configuration
const response = await supportAgent.handle({
  message: 'I never received my order #12345',
  context: {
    customer: {
      id: 'cus_789',
      name: 'Jane Smith',
      tier: 'Premium',
    },
  },
})
 
// Handle a complex business task with elegant configuration
const result = await operationsAgent.handle({
  task: 'Update customer information and notify the support team',
  context: {
    customer: customerData,
    changes: { email: 'new@example.com' },
    priority: 'High',
  },
})

Seamless Human Collaboration

Agents.do provides elegant handoffs between AI and human operators:

// Create an agent with human collaboration capabilities
const reviewAgent = new Agent({
  name: 'content-reviewer',
  description: 'Reviews and improves marketing content',
  humanCollaboration: {
    enabled: true,
    escalationThreshold: 0.7, // Confidence threshold for escalation
    assignTo: 'marketing-team',
  },
})
 
// Handle a task with potential human handoff
const reviewResult = await reviewAgent.handle({
  content: draftContent,
  guidelines: brandGuidelines,
  deadline: '2023-04-10T15:00:00Z',
})
 
// Check if the task was handled by AI or escalated
if (reviewResult.handledBy === 'ai') {
  console.log('AI completed the review:', reviewResult.improvements)
} else {
  console.log('Task escalated to:', reviewResult.assignedTo)
  console.log('Escalation reason:', reviewResult.escalationReason)
}

The .do  Ecosystem

Agents.do is designed to work seamlessly with other .do  services:

Installation

npm install agents.do
# or
yarn add agents.do
# or
pnpm add agents.do

Quick Start

import { Agent } from 'agents.do'
 
// Create a customer support agent with minimal configuration
const supportAgent = new Agent({
  name: 'customer-support',
  description: 'Assists customers with product questions and technical issues',
})
 
// Use the agent to handle a customer inquiry
const response = await supportAgent.handle({
  message: "I'm having trouble connecting my device to WiFi",
  context: {
    customer: {
      id: 'cus_12345',
      name: 'John Doe',
      recentPurchases: [{ id: 'prod_789', name: 'Smart Hub' }],
    },
  },
})
 
console.log(response.message)
// "I see you recently purchased a Smart Hub. Let me help you connect it to WiFi..."
Last updated on
Manifesto