Agentic Patterns

6 points by nkko 2 days ago

I've put together a catalog of repeatable patterns for building effective AI agents: https://agentic-patterns.com

This is a very early version - many entries are AI-assisted placeholder content that needs refinement from real practitioners. The goal is to collect the practical techniques teams keep independently rediscovering (context window management, sub-agent spawning, feedback loops, etc).

If you've built agents and recognized patterns that made them more effective, I'd love your contributions.

http://github.com/nibzard/awesome-agentic-patterns/

The criteria are simple:

Repeatable - more than one team is using it Agent-centric - improves how an AI agent senses, reasons, or acts Traceable - backed by a public reference

Inspired by the "Raising an Agent" podcast from Sourcegraph. PRs welcome for both new patterns and improvements to existing ones.