OpenAI's official lightweight multi-agent Python SDK
openai-agents is OpenAI's official MIT Python SDK: Agent + Runner, with tracing and handoffs. It is not ChatGPT, and not a terminal coding agent.
Updated 2026-09-04. Same-day snapshot of the official repo.
How to Get Started with the OpenAI Agents SDK
3 steps to one Agent from the official Quickstart.
Install
Official docs: pip or uv after a venv.
pip install openai-agentsSet the API key
Export OPENAI_API_KEY for this shell. Don't commit the key.
export OPENAI_API_KEY=...Run an Agent
Official Quickstart uses Agent + Runner. Full async example is in the docs.
from agents import Agent, RunnerPitfalls
This is not a ChatGPT hub
ChatGPT is a chat product; this hub is a Python SDK. Model brands live under ai-brands nav, not here.
Don't skip tracing / guardrails in production
Official docs treat tracing, sessions and guardrails as first-class. PoC can start with Runner.run; production needs to see each step.
Package name ≠ import path
You install `openai-agents` but import `from agents import ...`. Installing `openai` alone is not this SDK.
Beside LangChain & Claude Code
Based on official docs and licenses. This hub is an SDK — not ChatGPT, and not a terminal coding agent.
| Tool | License | Surface | Positioning | Best for |
|---|---|---|---|---|
| OpenAI Agents SDK | MIT | Python SDK | Thin multi-agent + tracing | Production agents already on OpenAI |
| LangChain | MIT | Python / JS framework | Components and many integrations | Mixing many models and tools |
| Claude Code | Proprietary / closed | Terminal agent product | Edit code in a repo | Day-to-day coding, not embedding an SDK |
Official Directory
Official
- GitHub repository
Official Python SDK. MIT.
Docs & Install
- Documentation
Concepts, Runner, handoffs, sessions.
- Quickstart
venv, pip install, first Agent.
- PyPI · openai-agents
`pip install openai-agents`.
Related
- OpenAI platform
API keys and models — not this SDK's docs.
Latest Updates
2026-09-04 · GitHub · openai/openai-agents-python
OpenAI Agents SDK: lightweight multi-agent workflowsOfficial Python package openai-agents, MIT. Docs centre on Agent + Runner, plus tracing / handoffs. ~29k stars.
LaoJin's Pick
FAQ
- What is the OpenAI Agents SDK?
- OpenAI's official lightweight Python framework for multi-agent workflows. Package `openai-agents`, import `from agents import Agent, Runner`. Not the ChatGPT web app.
- How do I install it?
- Official Quickstart: create a venv, `pip install openai-agents` (or `uv add openai-agents`), and set `OPENAI_API_KEY`.
- Does it lock me to OpenAI models?
- It defaults to an OpenAI API key. Official positioning is the shortest path in the OpenAI ecosystem; other providers are on you — it is not LangChain's integration catalog.
- Vs LangChain and Claude Code?
- Already on OpenAI and you want agents in code → this SDK. Many providers and retrieval components → LangChain. Editing a repo in a terminal → Claude Code / Gemini CLI, not an SDK.
Need to deploy a team-level Agent system?
Open-source frameworks are free, but turning a runtime into an enterprise asset requires access control, RBAC sandboxes, and private RAG integrations. Laojin Global provides commercial-grade architecture and deployment support.
Cite This Hub
We encourage you to capture this hub into your knowledge base, research, or feed it straight to an AI agent under CC BY 4.0.
How this hub stays current
Version and links live in src/content/openai-agents-sdk.ts; star snapshots sync via GitHub API.