The article presents a curated collection of agentic AI patterns aimed at enhancing the performance of autonomous or semi-autonomous AI agents in real-world applications. It highlights the gap between simplistic tutorials and the complexities of actual products, offering a resource that showcases practical, repeatable strategies. Each pattern listed is used by multiple teams and is backed by public references, ensuring that users can trust the information provided. Categories include Context & Memory, Feedback Loops, Learning & Adaptation, and Reliability & Evaluation, among others. Within these categories, you'll find specific methods such as episodic memory retrieval, self-healing retries, and proactive agent state externalization. The project encourages contributions, allowing users to suggest new patterns or improvements, emphasizing the collaborative nature of the resource. The list includes over a hundred patterns, each addressing different challenges in AI agent development, like tool use and orchestration. Some noteworthy entries include the "Graph of Thoughts" and "Agent Reinforcement Fine-Tuning," which focus on enhancing reasoning and learning capabilities. The project draws inspiration from earlier learnings in AI coding agents, making it a living document that evolves with community input.