An AI agent is an LLM-powered system that can autonomously decide which actions to take, call tools (web search, code execution, APIs), observe results, and iterate until it completes a goal. Unlike a simple chatbot, an agent can take multi-step actions without human input at each step.

What is the ReAct pattern in AI agents?

ReAct (Reasoning + Acting) is an agent architecture where the LLM alternates between Thought (reasoning about what to do next), Action (calling a tool), and Observation (receiving the tool result). This loop continues until the task is complete.

What is a multi-agent system?

A multi-agent system uses multiple specialized LLM agents working together. One orchestrator agent decomposes a task and delegates subtasks to worker agents with different tools or expertise. This enables parallelism and modularity for complex workflows.

What frameworks are used to build AI agents?

Popular frameworks include LangChain (most widely used), LangGraph (for stateful agent workflows), AutoGen (Microsoft's multi-agent framework), CrewAI (role-based agents), and the OpenAI Assistants API. Each has different trade-offs for production use.

Topic Hub

AI Agents

AI agents are LLMs that can reason, plan, use tools, and take multi-step actions to complete goals autonomously. Learn how they work, how to build them, and how to deploy them reliably in production.

What it is

What Is an AI Agent?

An AI agent is a software system where a language model acts as the reasoning engine and decision-maker. Given a goal, the agent can break the task into steps, call external tools (search, code execution, APIs, databases), observe the results, and iterate — all without a human directing each step.

The key difference from a chatbot is autonomy. A chatbot responds to a single message. An agent pursues a goal across multiple steps, adapting its plan based on tool outputs and new information.

ReAct

Alternates Thought → Action → Observation in a loop. The simplest and most widely used agent pattern.

Plan-and-Execute

Planner creates a full step-by-step plan first; executor carries it out. Better for long, structured tasks.

Multi-Agent

Orchestrator delegates to specialist sub-agents. Enables parallelism and separation of concerns.

Reflection

Agent critiques its own output and iterates until quality standards are met. Improves output reliability.

Memory

In-context, external (vector DB), or episodic memory lets agents recall past interactions and facts.

Tool Use

Function calling lets agents invoke code, APIs, search engines, browsers, and databases as needed.

Why it matters

Why AI Agents Are the Future of Software

Agents are shifting software development from "write code to automate tasks" to "describe goals and let AI execute them". The implications are significant:

Dramatically expanded automation — tasks that required complex deterministic code can be handled by a well-prompted agent with the right tools
Adaptability — agents handle edge cases by reasoning, not by explicit if-else logic
Human-in-the-loop workflows — agents can run autonomously but escalate to humans when uncertain
Composability — multi-agent systems can coordinate entire workflows across search, data processing, writing, and decision-making

Understanding agents is becoming as fundamental as understanding APIs or databases — it is the architectural pattern defining the next generation of AI-powered software.

Where it fits in the AI roadmap

Phase 5 of the AI Engineering Roadmap

AI Agents is Phase 5 of the AI roadmap for developers. Here is where it fits:

Phase 3 Prompt Engineering → required for tool-use and system prompts

Phase 4 RAG Systems → required for agent memory and retrieval

Phase 5 ★ AI Agents ← you are here

Phase 6 Fine-tuning → customise models to improve agent behaviour

Agents require mastery of both prompt engineering (for reasoning and tool-use schemas) and RAG (for memory). They are the highest-complexity topic in the standard AI engineering curriculum — but also the most powerful.

Tutorials on this site

AI Agent Guides & Tutorials

From agent fundamentals to multi-agent architectures — everything you need to build and deploy autonomous AI systems.

Guide AI Agent Fundamentals: How LLM Agents Think, Plan, and Act Guide Tool Use and Function Calling in LLMs: Practical Patterns and Best Practices Architecture Multi-Agent Systems: Architectures for Coordinating LLM Agents Tutorial LangChain Complete Tutorial: Chains, Agents, Tools, and Memory Project Building AI Chatbots: From Simple Q&A to Contextual Conversations Architecture AI Application Architecture: Patterns for Production AI Systems

Continue Learning

Hub Prompt Engineering — Techniques for writing effective LLM prompts Hub RAG Systems — Retrieval-Augmented Generation guides and tutorials Hub Fine-Tuning LLMs — Instruction tuning, LoRA, and custom model training Path AI Engineer Learning Path — Full structured curriculum

Ready to build your first AI agent?

Follow the complete AI engineering roadmap — structured phases from foundations to autonomous agents.

View the full AI roadmap →