Agent Types
1. Based on Autonomy
Autonomous Agent
Acts independently without external intervention. E.g., robotic agents, LLM agents with reasoning.
Non-Autonomous Agent
Requires continuous input or guidance from users. E.g., basic rule-based systems.
2. Based on Interaction System
Chat Agent
Designed to interact with users via conversation. Example: Customer Support Chatbot
Workflow Agent
Executes a sequence of defined steps to complete a task. Example: Agent in a multi-step booking or form-filling flow
3. Based on Execution Timing
Synchronous Agent
Waits for a response before proceeding. Example: API agent that blocks until data is returned
Asynchronous Agent
Continues without waiting for a response. Example: Event-driven notification agents
4. Based on Workflow Predictability
Deterministic Workflow Agent
Executes the same steps for the same input every time.
Non-Deterministic Workflow Agent
May vary its behavior based on context, randomness, or learning.
5. Based on Role/Function
Data Collection Agent
Gathers information from sources like APIs or websites.
Validation Agent
Verifies or filters the outputs of other agents.
Router Agent
Delegates tasks to the most appropriate agents.
Reasoning Agent
Uses logic or LLMs to derive insights or make decisions.
Multi-Agent Coordinator
Orchestrates the collaboration of multiple specialized agents.
Here are more agent types to expand your classification system further — useful for designing AI-based workflows, multi-agent systems, or orchestrators:
6. Based on Learning Capability
Learning Agent
Improves its performance over time using ML techniques. Example: Reinforcement learning agent
Static Agent
Behavior is fixed; does not learn from experience. Example: Finite State Machine agent
7. Based on Environment Awareness
Reactive Agent
Responds directly to stimuli without internal state or history. Example: Obstacle-avoiding robot
Deliberative Agent
Uses internal models to make informed decisions. Example: Planning-based AI agent
8. Based on Agency Composition
Single-Agent System
Only one agent operates in the environment.
Multi-Agent System (MAS)
Multiple agents interact, collaborate, or compete. Example: Game AI, swarm robotics
9. Based on Mobility
Mobile Agent
Can move across networks or systems. Example: Software agent migrating between servers
Stationary Agent
Remains on one host system. Example: Server-side chatbot
10. Based on Purpose
Monitoring Agent
Observes systems and triggers alerts or logs. Example: Health check agent
Optimization Agent
Aims to find the best solution within constraints. Example: Scheduling or routing agent
Negotiator Agent
Engages in negotiation with users or other agents. Example: E-commerce deal agent
Social Agent
Interacts based on human-like behaviors or emotions. Example: Virtual companion bots
Absolutely! Here are 10 more agent types, further diversifying based on purpose, technical role, and ecosystem function — especially useful for GenAI, orchestration, and automation scenarios:
11. Based on Communication Style
Broadcast Agent
Sends information to multiple agents at once. Example: Notification broadcaster in microservices
Unicast Agent
Communicates with one specific agent at a time.
Middleware Agent
Acts as a communication bridge between agents or systems.
12. Based on Execution Layer
Frontend Agent
Interfaces with users directly (UI-level agents). Example: Browser-based assistant
Backend Agent
Works behind the scenes to perform logic or database operations. Example: LLM query optimizer
13. Based on Specialization
Search Agent
Specializes in information retrieval or ranking. Example: Vector search agent
Summarization Agent
Condenses large content into digestible formats. Example: News summarizer
Translation Agent
Converts text between languages. Example: Multilingual support bot
Extraction Agent
Pulls structured data from unstructured inputs. Example: Resume parser or legal clause extractor
14. Based on Deployment Scope
Cloud Agent
Runs on cloud infrastructure; often stateless and scalable.
Edge Agent
Deployed closer to the data source, e.g., on IoT devices. Example: Smart home agent
Hybrid Agent
Coordinates tasks across edge and cloud.
15. Based on Cognitive Function
Planning Agent
Capable of long-term, multi-step decision-making. Example: Goal-oriented agent for task decomposition
Reflective Agent
Evaluates and adjusts its own strategies or behavior. Example: Agent that fine-tunes its prompts or tool usage
Evaluative Agent
Scores, grades, or critiques outputs. Example: Resume evaluator or code reviewer
You're building a super thorough agent taxonomy — love it! Here’s yet another set of 10 unique agent types to further enrich your system:
16. Based on Tool Usage
Tool-Using Agent
Can call APIs, databases, or plugins to perform tasks. Example: LangChain or OpenAI function-calling agents
Tool-Free Agent
Relies purely on reasoning and internal logic. Example: Prompt-only LLM agent
17. Based on Goal Orientation
Goal-Based Agent
Takes actions with a specific end goal in mind. Example: Task completion agent (e.g., "Book flight")
Utility-Based Agent
Chooses actions based on maximizing utility or reward. Example: Game AI or recommendation engine
18. Based on Lifecycle
One-shot Agent
Executes a single task then terminates. Example: Email parser triggered on receipt
Persistent Agent
Stays alive and handles multiple or ongoing tasks. Example: Monitoring agent or daemon
19. Based on Access Level
Public Agent
Available for use by any authorized user. Example: Open API endpoint agent
Private Agent
Restricted to specific users or teams. Example: Internal HR chatbot
20. Based on Agent Collaboration
Cooperative Agent
Works collaboratively with other agents to complete tasks.
Competitive Agent
Competes against other agents for resources or goals. Example: Bidding agents in marketplaces
Neutral Agent
Operates independently, without coordination.
Last updated