OpenAI Agents SDK

// Overview

OpenAI Agents SDK is a lightweight Python framework for building multi-agent workflows. You define agents as LLMs configured with instructions, tools, guardrails, and handoffs, then let them delegate work to each other to complete a task.

It is provider-agnostic: it works with the OpenAI Responses and Chat Completions APIs, plus 100+ other LLMs. Built-in sessions handle conversation history across runs, and tracing lets you view, debug, and optimize your workflows.

It fits the multi-agent systems space by keeping the core small and explicit. Instead of a heavy orchestration layer, you compose agents, tools, and handoffs directly in Python, which suits developers who want control over how agents coordinate.

// What it does

• Agents: LLMs configured with instructions, tools, guardrails, and handoffs
• Handoffs and agents-as-tools for delegating specific tasks to other agents
• Tools support: Python functions, MCP, and hosted tools
• Guardrails for input and output validation, plus human-in-the-loop controls
• Sessions for automatic conversation history, and built-in tracing for debugging
• Provider-agnostic across OpenAI Responses, Chat Completions, and 100+ other LLMs

// Getting started

Python 3.10 or newer is required. Create a virtual environment, install the package, and run an agent.

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install openai-agents

uv init
uv add openai-agents

export OPENAI_API_KEY=sk-...

from agents import Runner
from agents.run import RunConfig
from agents.sandbox import Manifest, SandboxAgent, SandboxRunConfig
from agents.sandbox.entries import GitRepo
from agents.sandbox.sandboxes import UnixLocalSandboxClient

agent = SandboxAgent(
    name="Workspace Assistant",
    instructions="Inspect the sandbox workspace before answering.",
    default_manifest=Manifest(
        entries={
            "repo": GitRepo(repo="openai/openai-agents-python", ref="main"),
        }
    ),
)

result = Runner.run_sync(
    agent,
    "Inspect the repo README and summarize what this project does.",
    run_config=RunConfig(sandbox=SandboxRunConfig(client=UnixLocalSandboxClient())),
)
print(result.final_output)

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Coordinate several specialized agents that hand off tasks to each other in a workflow
• Give an agent tools (Python functions, MCP, or hosted tools) to take real actions
• Build voice agents with realtime models using full agent features
• Add guardrails and human-in-the-loop checks to validate agent input and output

// How OpenAI Agents SDK compares

OpenAI Agents SDK alongside other open-source multi-agent systems tools AI/TLDR tracks, ranked by GitHub stars.