Working through this module?0 of 19 done
Module 03 · Page 4 of 20 · 8 sections

Types of Agents

Reactive, deliberative, hybrid, tool-augmented, multi-agent and hierarchical agents compared.

02Agent Architecture and Components08Multi-Agent Systems in Agentic AI14Agentic AI Design Patterns and Architectures
3.1

Overview

  • Agentic AI has different types of agents based on their architecture, capabilities and use cases.
  • Each type of agent is designed to handle specific tasks and complexity.
  • In real-world applications, multiple types of agents are often combined together.
3.2

Classification of Agents

Types of Agents
Based on Architecture
  • Single Agent
  • Multi-Agent
  • Hierarchical
  • Hybrid
Based on Reasoning
  • Reactive
  • Deliberative
  • Hybrid
Based on Tool Usage
  • Tool-Augmented
  • RAG-based
  • Environment aware
Based on Collaboration
  • Cooperative
  • Competitive
  • Human-in-the-Loop
Four lenses for classifying an agent.
3.3

Reactive Agent

  • Responds to the current environment without maintaining a long-term plan or memory.
  • Uses simple rules or policies.
  • Works well for simple and time-sensitive tasks.
  • Example: chatbot, real-time alert system.
EnvironmentInput
Reactive AgentRule based
ActionOutput
3.4

Deliberative Agent

  • Plans and reasons before taking action.
  • Uses internal models and long-term goals.
  • Handles complex tasks by considering future outcomes.
  • Example: research assistant, task planning agent.
GoalInput
Planning & Reasoning (LLM)
Internal ModelWorld knowledge + memory
ActionOutput
Related05Planning and Reasoning in Agentic AI
3.5

Hybrid Agent

  • Combines reactive and deliberative approaches.
  • Uses quick responses for simple tasks and planning for complex tasks.
  • Example: personal assistant, autonomous workflow agent.
EnvironmentInput
Reactive ModuleFast response
Deliberative ModulePlanning & reasoning
ActionOutput
3.6

Tool-Augmented Agent

  • Uses external tools, APIs and systems to perform tasks.
  • Can search the web, query databases, run code, access files and use third-party services.
  • Greatly extends the capabilities of LLMs.
  • Example: travel planner, data analyst agent.
User Query
AgentLLM
Web Search
Database
Code Execution
APIs / Apps
Final Response
Related06Tool Use and Function Calling in Agentic AI
3.7

Multi-Agent System

  • Multiple agents work together to solve complex tasks.
  • Each agent has a specific role (e.g. planner, researcher, executor, reviewer).
  • Agents can communicate and collaborate.
  • Example: research workflow, content generation pipeline.
User Query
Manager AgentOrchestrator
Research AgentFind information
Writer AgentGenerate content
Reviewer AgentCheck & refine
Final Output
Related08Multi-Agent Systems in Agentic AI
3.8

Hierarchical Agent

  • Organized in a hierarchy with a manager agent and sub-agents.
  • Manager agent breaks down the task and assigns it to sub-agents.
  • Useful for large and complex workflows.
  • Example: enterprise automation, software development.
Manager AgentTask planning & delegation
Sub-Agent 1Research
Sub-Agent 2Analysis
Sub-Agent 3Execution
Final Result