awesome-agentic-patterns

Why we included this project

Most agent tutorials stop at the toy demo, so teams hitting real production failures are left to reinvent the same fixes. This catalogue collects the patterns that actually show up in working agents, from context-window management and feedback loops to guardrails and tool-use tricks, and every entry links back to a public blog post, talk, repo, or paper. Each pattern is documented with the mechanism, when it applies, when to skip it, and known limitations, which is the difference between a link dump and something a team can actually decide from. The companion website adds filtering, side-by-side comparison, and a decision guide, so the material is navigable rather than a wall of README. It is a reference to learn from, not software to deploy, and that is exactly what makes it useful as a shared vocabulary for designing and reviewing agent architectures.