Disclaimer: AI was used extensively to generate the content of this project. That decision allowed for quick and comprehensive coverage of the topics. The goal of the project is to support understanding of agentic AI design patterns and provide an opportunity for collaboration and thought-sharing on these ideas.

Introduction

In the rapidly evolving field of artificial intelligence, we are witnessing the emergence of a new class of applications—agentic AI systems that can reason, plan, and take actions to accomplish complex tasks. Just as object-oriented software development matured through the identification and codification of design patterns, the field of agentic AI is now reaching a stage where recurring architectural and behavioral patterns are becoming apparent and necessary.

This project aims to catalog, describe, and analyze these emerging design patterns for agentic AI applications. Drawing inspiration from the seminal work "Design Patterns: Elements of Reusable Object-Oriented Software" by Gamma, Helm, Johnson, and Vlissides (the "Gang of Four"), we have adopted a similarly structured approach to documenting these patterns. Our goal is to provide AI practitioners, researchers, and developers with a common vocabulary and a set of proven solutions to recurring problems in the design and implementation of agentic systems.

See the Overview of Patterns page for a list of pattern categories and links to individual design patterns.

Why Design Patterns for Agentic AI?

Agentic AI applications present unique challenges that traditional software design patterns do not fully address. These systems must balance autonomy with safety, handle uncertainty and incomplete information, collaborate effectively with humans, and maintain transparency in their decision-making processes. The patterns in this book emerge from practical experience building and deploying such systems, offering insights into how to tackle these challenges effectively.

The Collaborative Nature of This Project

This project is conceived as a living document that will evolve through the contributions of the AI community. As agentic AI continues to advance, new patterns will emerge, and existing patterns will be refined. We invite practitioners to contribute their experiences, insights, and critiques to help build a comprehensive resource that reflects the collective wisdom of the field.

Design pattern samples are managed in the accompanying repo.

Pattern Definition Template

Each pattern in this collection follows a consistent format defined by the content below:

The pattern name will be indicated in the page title.

Classification

[Indicate whether the pattern is Core Processing, Architectural, Efficiency, Memory Management, Safety, Human Collaboration, Orchestration, Explainability, or Resilience pattern]

Intent

A concise statement describing what the pattern does and what problem it aims to solve.

Also Known As

Alternative names for the pattern, if any exist.

Motivation

A real-world scenario explaining why this pattern is useful, including:

• The problem or challenge it addresses
• Why traditional approaches may be insufficient
• A diagram or visual representation where appropriate

Applicability

When to use this pattern:

• Specific scenarios where this pattern should be applied
• Indicators that suggest this pattern would be beneficial
• Prerequisites for successful implementation

Structure

A visual representation of the pattern showing:

• Components and their relationships
• Data and control flow
• Key interfaces and interactions

Components

The key elements participating in the pattern:

• [Component Name]: Description of its role and responsibilities
• [Component Name]: Description of its role and responsibilities
• ...

Interactions

How the components work together:

• Sequence of operations
• Communication patterns
• Coordination mechanisms

Consequences

The results and trade-offs of using the pattern:

• Benefits: Positive outcomes and advantages
• Limitations: Constraints and potential drawbacks
• Performance implications: Effects on system resources and response times

Implementation

Guidelines for implementing the pattern:

• Step-by-step approach
• Key considerations and decision points
• Language-specific or framework-specific tips
• Common pitfalls and how to avoid them

Code Examples

Sample implementations showing how to realize the pattern:

# Python implementation example

Variations

Common modifications or adaptations of the basic pattern:

• [Variation Name]: Description and use cases
• [Variation Name]: Description and use cases

Real-World Examples

Systems or applications where this pattern has been successfully applied:

• Example 1: Description and outcomes
• Example 2: Description and outcomes

Related Patterns

Other patterns that:

• Are frequently used with this pattern
• Serve similar purposes but in different contexts
• Complement or extend this pattern

All Patterns

Go to the Overview of Patterns page to see an overview of the categories of the design patterns.