Agentic Patterns (github.com)

A new curated catalogue titled "Agentic Patterns" has been launched, showcasing a collection of real-world techniques and workflows that enhance the capabilities of autonomous and semi-autonomous AI agents. This initiative aims to bridge the gap between theoretical tutorials and practical applications in production environments, highlighting tested patterns that multiple teams have successfully integrated. The catalogue categorizes various patterns into areas such as orchestration, context, feedback loops, and reliability, providing a resource for AI developers seeking to build better-performing agents. The significance of this project lies in its agent-centric approach, focusing on improving how AI systems sense, reason, and act in their environments. By surfacing repeatable practices that are adequately referenced and backed by public documentation, the catalogue promotes a community-driven effort to refine agent development. Among the notable patterns included are techniques for memory management, task decomposition, and multi-agent orchestration, which provide foundational insights for building more reliable and efficient AI systems. This resource not only empowers current AI projects but also encourages ongoing innovation and collaboration in the AI/ML community.