Agents.do

AI agents are autonomous software entities that perceive their environment, make decisions, and take actions to achieve specific goals. On the .do platform, agents transform manual work into automated, scalable operations.

What Are Agents?

Agents combine several key capabilities:

Perception - Understand context through natural language, APIs, and data
Reasoning - Make decisions using LLMs, logic, and learned patterns
Action - Execute tasks through API calls, code generation, and tool use
Learning - Improve performance through feedback and experience
Autonomy - Operate independently within defined guardrails

Agent Architecture

Core Components

Agent Runtime

Event-driven execution model
State management and memory
Tool and API integration
Safety constraints and approval workflows

Agent Definition

Role and responsibilities
Available tools and capabilities
Decision-making logic
Success criteria

Agent Coordination

Multi-agent orchestration
Task delegation and routing
Shared context and memory
Team collaboration patterns

Types of Agents

Task Agents

Single-purpose agents that execute specific tasks:

Data extraction and transformation
Content generation and editing
API integrations and automations
Monitoring and alerting

Workflow Agents

Multi-step agents that orchestrate complex processes:

Order fulfillment pipelines
Customer onboarding flows
Content approval workflows
Financial reconciliation

Autonomous Agents

Self-directed agents that pursue long-term goals:

Business development and sales
Product management and planning
Customer success management
Continuous improvement

Platform Integration

Agents integrate seamlessly with platform primitives:

import { $, on, send, ai } from 'sdk.do'

// Define agent
const salesAgent = $.Agent.create({
  name: 'Sales Development Agent',
  role: 'Generate and qualify leads',
  capabilities: ['research', 'outreach', 'qualification'],
  tools: ['linkedin-api', 'email', 'crm'],
})

// Event-driven execution
on.Lead.created(async (lead) => {
  const research = await ai.research({ company: lead.company })
  const qualified = await salesAgent.qualify({ lead, research })

  if (qualified.score > 0.7) {
    send.Lead.qualified(qualified)
  }
})

Agent Development

Creating Agents

Define Role - What is the agent responsible for?
Specify Tools - What APIs and capabilities does it need?
Implement Logic - How does it make decisions?
Set Guardrails - What are the safety constraints?
Test & Deploy - Validate behavior and monitor performance

Best Practices

Start with narrow, well-defined tasks
Provide clear success criteria
Implement human-in-the-loop for high-stakes decisions
Monitor performance and costs continuously
Version agent definitions for reproducibility

Next Steps

Overview - Deep dive into agent concepts
Capabilities - Explore agent skills and tools
Teams - Learn about multi-agent collaboration
Services - Turn agents into sellable services
Examples - See real-world agent implementations

Agents.do

On this page