Skip to main content

​
Core Concepts Overview

Erdo is built around a few key concepts that work together to create powerful AI automation workflows. Understanding these concepts will help you build effective agents and leverage the full potential of the platform.

​
Platform Architecture

Agents

The main containers for AI workflows - define what your automation does

Steps

Individual actions within agents - the building blocks of workflows

Actions

Specific operations like LLM calls, code execution, or web searches

Result Handlers

Conditional logic that responds to step outcomes and controls workflow flow

Tools

External integrations that extend agent capabilities

​
How It All Works Together

​
1. Agents

Agents are the top-level containers that define complete workflows. They:
  • Orchestrate multiple steps
  • Define execution order and dependencies
  • Handle errors and retries
  • Manage resources and permissions

​
2. Steps

Steps are the individual actions within an agent. Each step:
  • Performs one specific operation
  • Can depend on previous steps
  • Has configurable execution conditions
  • Produces outputs for subsequent steps

​
3. Actions

Actions define what each step actually does:
  • LLM Actions: Generate text, analyze content, make decisions
  • Code Execution: Run Python code for data processing
  • Memory Actions: Store and retrieve knowledge
  • Web Research: Search and parse web content
  • Utility Actions: Data transformation and flow control

​
4. Result Handlers

Result Handlers add conditional logic to your workflows:
  • Conditional Execution: Run different steps based on outcomes
  • Error Recovery: Handle failures and implement retry logic
  • Dynamic Routing: Branch workflows based on data or results
  • Success Actions: Execute follow-up steps when conditions are met

​
5. Tools & Integrations

Tools extend agent capabilities:
  • Web search and content parsing
  • API integrations
  • Database connections
  • File processing
  • Custom functions

​
Key Features

  • Conditional execution based on data and results
  • Dynamic branching and decision making
  • Parallel processing for efficiency
  • Automatic retry and error handling
  • Persistent knowledge storage across executions - Semantic search for relevant information - Organizational knowledge sharing - Continuous learning from interactions
  • Sandboxed code execution environment - Role-based access controls - Secure credential management - Input validation and safety controls
  • Auto-scaling based on workload
  • Resource limits and optimization
  • Monitoring and alerting
  • Production-ready deployment

​
Common Workflow Patterns

​
Data Analysis Pipeline

Input Data → Validation → Processing → Analysis → Insights → Storage

​
Research Automation

Query → Web Search → Content Analysis → Synthesis → Report Generation

​
Business Process Automation

Trigger Event → Data Collection → Decision Logic → Action Execution → Notification

​
Content Processing

Raw Content → Classification → Extraction → Transformation → Distribution

​
Getting Started Path

1

Understand Agents

Learn how agents work and what makes them powerful
2

Master Steps

Understand how to build workflows with individual steps
3

Explore Actions

Discover all the actions available for your steps
4

Build Your First Agent

Create a simple automation to see concepts in action

​
Best Practices

  • Single Responsibility: Each agent should have one clear purpose - Modular Steps: Break complex logic into smaller, reusable steps - Error Handling: Always plan for failure scenarios - Documentation: Clearly document agent purpose and usage

​
Next Steps

Ready to dive deeper? Here’s where to go next:

Build Your First Agent

Follow our quick start guide to create your first agent

Explore Examples

See real-world examples of agents in action

SDK Reference

Complete reference for building agents with code

CLI Tools

Learn the command-line tools for development