What are all the types of AI agents?

Question

Accepted Answer

## Types of AI Agents

AI agents can be classified by architecture, capability, and role. Here are the major types you'll encounter in Gen AI engineering.

### By Architecture

| Type | Description | Example |
|------|-------------|---------|
| **Simple Reflex Agent** | Reacts to current input only, no memory | Spam filter |
| **Model-Based Agent** | Maintains internal world model | Robot navigation |
| **Goal-Based Agent** | Acts to achieve defined goals | Task planner |
| **Utility-Based Agent** | Maximizes a utility/reward function | Game AI |
| **Learning Agent** | Improves over time from experience | RLHF-tuned LLM |

### By Role in Multi-Agent Systems

| Type | Description |
|------|-------------|
| **Orchestrator** | Plans and delegates tasks to other agents |
| **Sub-agent / Worker** | Executes specific tasks given by orchestrator |
| **Critic / Reviewer** | Evaluates and provides feedback on other agents' output |
| **Tool-use Agent** | Focuses on calling external tools (search, code, APIs) |
| **Memory Agent** | Manages retrieval and storage in long-term memory |

### By Capability / Modality

| Type | Input/Output | Example |
|------|-------------|---------|
| **Language Agent** | Text in, text out | ChatGPT-style assistant |
| **Code Agent** | Natural language → code + execution | Devin, GitHub Copilot agent |
| **Vision Agent** | Image + text in, text out | GPT-4V with tools |
| **Multimodal Agent** | Text + image + audio | Claude 3.5 with vision |
| **Embodied Agent** | Sensor input → physical action | Robotics AI |
| **Web Agent** | Browser control | Playwright MCP, Browser Use |

### By Execution Pattern

| Pattern | How It Works |
|---------|-------------|
| **ReAct Agent** | Reason → Act → Observe loop |
| **Plan-and-Execute** | Create full plan, then execute each step |
| **Reflexion** | Self-critique and retry on failure |
| **MRKL** | Routes to specialized modules based on query |
| **Tree of Thoughts** | Explores multiple reasoning paths |

### By Autonomy Level

```
Low autonomy ──────────────────────────── High autonomy
     │                                          │
  Chatbot          Tool-user        Autonomous agent
  (responds)    (calls APIs)        (plans + acts)
```

### Popular Framework Implementations

| Framework | Agent Types Supported |
|-----------|----------------------|
| **LangGraph** | Stateful graph-based agents |
| **AutoGen** | Conversational multi-agent |
| **CrewAI** | Role-based agent crews |
| **Google ADK** | Hierarchical agent orchestration |
| **Claude + tools** | Tool-use and agentic Claude |

What are all the types of AI agents?

Answer

Types of AI Agents

By Architecture

By Role in Multi-Agent Systems

By Capability / Modality

By Execution Pattern

By Autonomy Level

Popular Framework Implementations

Related Concepts

What is AI?

What are all the current types of AI?

What is Machine Learning (ML)?

What is Deep Learning in AI?

What is an LLM?

Type	Description	Example
Simple Reflex Agent	Reacts to current input only, no memory	Spam filter
Model-Based Agent	Maintains internal world model	Robot navigation
Goal-Based Agent	Acts to achieve defined goals	Task planner
Utility-Based Agent	Maximizes a utility/reward function	Game AI
Learning Agent	Improves over time from experience	RLHF-tuned LLM

Type	Description
Orchestrator	Plans and delegates tasks to other agents
Sub-agent / Worker	Executes specific tasks given by orchestrator
Critic / Reviewer	Evaluates and provides feedback on other agents' output
Tool-use Agent	Focuses on calling external tools (search, code, APIs)
Memory Agent	Manages retrieval and storage in long-term memory

Type	Input/Output	Example
Language Agent	Text in, text out	ChatGPT-style assistant
Code Agent	Natural language → code + execution	Devin, GitHub Copilot agent
Vision Agent	Image + text in, text out	GPT-4V with tools
Multimodal Agent	Text + image + audio	Claude 3.5 with vision
Embodied Agent	Sensor input → physical action	Robotics AI
Web Agent	Browser control	Playwright MCP, Browser Use

Pattern	How It Works
ReAct Agent	Reason → Act → Observe loop
Plan-and-Execute	Create full plan, then execute each step
Reflexion	Self-critique and retry on failure
MRKL	Routes to specialized modules based on query
Tree of Thoughts	Explores multiple reasoning paths

Framework	Agent Types Supported
LangGraph	Stateful graph-based agents
AutoGen	Conversational multi-agent
CrewAI	Role-based agent crews
Google ADK	Hierarchical agent orchestration
Claude + tools	Tool-use and agentic Claude