openai-agents-python

Why we included this project

Most LLM apps start with a single prompt, but once you need an agent to call tools, pass work to another agent, or run for a long time, you want a framework that handles the plumbing. The OpenAI Agents SDK keeps that plumbing small: agents, tools, handoffs, guardrails, and sessions are the primitives, and you compose them into multi-step workflows where one agent can delegate to another or call functions and MCP servers. It works with OpenAI's Responses and Chat Completions APIs as well as 100+ other LLMs, so you are not locked into one model provider, and built-in tracing shows you what each agent actually did during a run. Voice and realtime agents are supported out of the box, and sandbox agents can keep working inside a container over long tasks. The docs and examples are solid enough that a small team can go from install to a working multi-agent pipeline in an afternoon.