Skip to main content
Agents are systems that intelligently accomplish tasks—from simple goals to complex, open-ended workflows. This guide explains the fundamental concepts you need to understand, regardless of which framework you choose.

​
Quick Start

Create your first agent in three steps:
  1. Create a model provider - Configure your LLM connection
  2. Create an agent - Define your agent’s behavior and capabilities
  3. Run the agent - Execute conversations and get responses
For a complete step-by-step guide, see the Quickstart tutorial.

​
Agent Architecture

Agents follow a consistent architecture pattern across all frameworks:

​
Core Components

​
Agent Definition

Agents are created with a configuration that defines their identity, capabilities, and behavior.

Agent Configuration

Learn how to configure agents with providers, instructions, and tools

​
Memory Management

Built-in memory management with configurable limits and cleanup strategies.

Memory Management

Configure memory limits and cleanup strategies

​
Tools

Tools extend your Agent’s capabilities by providing access to external APIs, databases, and custom functions.

Tools

Learn how to create and use tools with your agents

​
Agent Execution

​
Execution Models

Different frameworks support different execution models, from simple handlers to complex graph-based workflows.

Running Agents

Learn how to execute agents and handle responses

​
Workflow Patterns

Agents support various workflow patterns including linear execution, conditional routing, and human-in-the-loop interactions.

Human-in-the-Loop

Implement interactive workflows with user input

​
Advanced Features

​
Deployment & Integration

​
Server Deployment

Deploy your Agents as HTTP servers with automatic RESTful endpoints.

Server Configuration

Deploy agents as production-ready HTTP servers

​
Frontend Integration

Integrate agents into your applications with client SDKs and UI components.

Client SDKs

Integrate agents into your frontend applications

​
Best Practices

Begin with basic agents and gradually add complexity. Use simple execution patterns for development and more advanced patterns for production.
Define custom state structures with proper typing to catch errors early and improve development experience.
Use schema validation for tools and provide clear descriptions for better LLM understanding.
Always implement proper error handling to provide good user experience even when things go wrong.
Use client SDKs and UI components for consistent user experiences across applications.

​
Next Steps

Quickstart

Get started with building agents

Tools

Add capabilities with tools

Human-in-the-Loop

Implement complex control flows

Server Integration

Deploy your Agents