OpenAI Agents SDK
Python SDK for building multi-agent LLM workflows with tools and handoffs
What it is
A Python library where each agent is an LLM given instructions, tools, guardrails, and the ability to hand off work to other agents, which can then run in text, voice, realtime, or sandboxed container modes. It works with OpenAI's Responses and Chat Completions APIs and many other model providers, and includes sessions for conversation history, human-in-the-loop approvals, MCP tool support, and built-in tracing. Developers building LLM-powered assistants and automated workflows use it to wire up and debug agent runs in code.