Framework · OpenAI
OpenAI Agents SDK at a glance
OpenAI's open-source Python SDK for agents, kept deliberately thin, with handoffs, guardrails, sessions and tracing built in. It defaults to OpenAI models but can call other providers, and gets durable runs through a Temporal integration that went GA in March 2026.
Best for: Minimal Python agents on OpenAI models that need handoffs, guardrails, sandboxed workspaces and tracing from day one.
Key facts
- Maintainer
- OpenAI
- Languages
- Python
- Licence
- MIT
- Latest version
0.22.3· September 17, 2026- Repository
- openai/openai-agents-python
- Hosted option
- OpenAI Agent Builder (AgentKit) + ChatKit
Quickstart
Install the package, set the credentials, then run the smallest agent from the official docs. Model IDs are the ones the docs use and may not be enabled on your account.
pip install openai-agentsfrom agents import Agent, Runner
agent = Agent(name="Assistant", instructions="You are a helpful assistant")
result = Runner.run_sync(agent, "Write a haiku about recursion in programming.")
print(result.final_output)Features
| Any model | Yes | OpenAI by default; Any-LLM and LiteLLM adapters are beta |
|---|---|---|
| MCP client | Yes | |
| MCP server | Not confirmed | |
| Multi-agent | Yes | Handoffs and agents-as-tools |
| Durable runs | Yes | Temporal integration via OpenAIAgentsPlugin, plus resumable run state |
| Human approval | Yes | Interruptions that pause a run and resume from saved state |
| Tracing | Yes | |
| Evals | No | No eval runner; OpenAI's Evals product closes 30 Nov 2026 |
Recent changes
- March 2026: the Temporal integration reached GA, turning model calls into durable Activities.
- June 2026: OpenAI announced that Agent Builder and its Evals product shut down on 30 November 2026.
- Sandbox agents and realtime voice agents were added.