Why this knowledge base exists
AI agents are moving from single responses to long-running systems that plan, use tools, inspect results, and continue working. Reliable outcomes depend on the operating loop around the model—not on a clever prompt alone.
What we cover
We document objectives, context, harnesses, evaluators, memory, budgets, stop conditions, observability, guardrails, recovery, and human approval. Closely related practices are included when they clarify how a complete loop operates.
Independence
Loop Engineering is an independent knowledge base. It is not affiliated with OpenAI, Anthropic, Google, or any other model or tooling provider.