Without API documentation, LLM agents hallucinate endpoints, invent parameters, and guess auth flows -- scoring just 0.399 accuracy in blind tests. LAP fixes this. One command gives your agent a ...