LangChain popularized the term in April 2026 as a multi-agent coordination model where agents act as digital team members with defined roles, shared memory, and observability across the full software delivery pipeline — not merely faster code generation. Zhenfeng Cao's June 2026 arXiv paper formalizes it against traditional software engineering: decision logic moves from pre-written code to runtime LLM reasoning, and humans shift to intent architects, coordinators, and outcome auditors. It sits alongside loop engineering (single-agent cycles) and graph engineering (multi-agent org wiring) as the orchestration layer practitioners build on top of.