Comparisons

No named team on record runs both. Pick CrewAI for role-based crews on any vendor's models with a managed platform, as PwC and Gelato describe; pick the OpenAI Agents SDK for a small loop on OpenAI models whose traces feed your evals, as Harvey's do.

CrewAI starts from role-based crews inside event-driven flows, with a managed visual builder. LangGraph has your team write down every step and transition, with state checkpointed between them. Both are MIT-licensed Python frameworks for multi-step agents.