OpenAI Agents SDK

The production-ready SDK for building complex, multi-agent AI workflows with robust orchestration, guardrails, and built-in observability.

The OpenAI Agents SDK is a lightweight, powerful framework for developing agentic AI applications, serving as a production-ready evolution of the previous Swarm experimentation. It’s a Python-first, provider-agnostic tool, supporting the OpenAI Responses API and 100+ other LLMs. Core primitives—Agents, Handoffs, Guardrails, and Sessions—enable sophisticated coordination: Agents use tools, Handoffs delegate tasks between specialized agents, Guardrails validate inputs and outputs, and Sessions manage conversation history automatically. The SDK includes built-in tracing, crucial for visualizing, debugging, and monitoring complex agent flows, ensuring you build reliable, real-world applications efficiently.

https://openai.github.io/openai-agents-python/