nibzard/awesome-agentic-patterns

What it solves

This project addresses the gap between simple AI tutorials and the complex, messy reality of building production-ready AI agents. It provides a curated library of proven, repeatable patterns that help developers move beyond basic demos to build autonomous or semi-autonomous agents that actually get work done.

How it works

The project acts as a catalogue of "agentic patterns"—specific workflows, mini-architectures, and techniques. Each pattern is vetted to ensure it is repeatable across different teams, agent-centric (improving how an agent senses, reasons, or acts), and traceable to public references like blog posts, papers, or repositories. The patterns are organized into categories such as memory management, feedback loops, orchestration, and security.

Who it’s for

It is designed for developers and engineers who are building or scaling autonomous AI agents and need practical, battle-tested architectural solutions to common production challenges.

Highlights

• Extensive Categorization: Browse patterns across eight key areas, including Context & Memory, Reliability & Eval, and Security & Safety.
• Interactive Discovery: The companion website offers a Pattern Explorer, a decision guide for selecting the right pattern, and visual maps of how patterns relate to one another.
• Evidence-Based: Every pattern is backed by real-world references, ensuring they are not just theoretical ideas but proven methods.
• Community-Driven: The catalogue is fluid and open to contributions, allowing the community to refine and expand the collection of patterns as the field evolves.