ToolsCompaniesAbout

OpenAI Agents SDK

by OpenAI@openaiSan Francisco, California, USANo ratings yet

Open-source Python framework for building multi-agent apps with tools, handoffs and guardrails.

Visit openai.github.io
Tool type
Open-source agent SDK
Made for
Developers
Runs on
Your own Python environment
Runs in your application; the sandbox example uses a local Unix client on macOS/Linux, with Docker or hosted clients for Windows.
Last checked
28 Oct 2026
5 product sources

What it does

The OpenAI Agents SDK is a code library, not an app you open in a browser. You write Python (a JavaScript/TypeScript version also exists) that defines agents: language models given instructions, tools and safety checks. You run them with a runner, and the SDK manages the loop until the task completes. You get back a result with the final output, run history and resumable state.

Developers use it to wire agents together so one can hand work to another or act as a tool for a manager agent. It also covers function tools generated from Python functions, MCP tool calling, sessions for memory, human-in-the-loop pauses and built-in tracing for debugging. It additionally offers sandbox agents that work in isolated file workspaces, plus realtime and voice agents. It defaults to the Responses API for OpenAI models but is described as provider-agnostic.

It suits developers who want a small set of primitives while keeping control of their own server, storage and tool code. It requires Python 3.10 or newer, installation with pip, and an OPENAI_API_KEY when using OpenAI models. The SDK is MIT-licensed and open source, but model calls, hosted tools and containers are billed separately by the API. It is not suited to non-programmers, and teams wanting a managed hosted harness should compare OpenAI's separate Agents API.

What it can do

  • Define agents with instructions, tools, guardrails and a built-in run loop
  • Delegate between agents using handoffs or agents-as-tools
  • Turn Python functions into tools and call remote MCP tools
  • Run sandbox agents in isolated workspaces with resumable sessions
  • Build realtime and speech-to-text/text-to-speech voice agents
  • Trace, debug and pause runs for human approval

Works with

  • MCP servers
  • OpenAI Responses API
  • OpenAI Chat Completions API
  • Redis (session storage, optional)

Example workflows

Triage with specialist agents
Input: customer questions and a few Python functions as tools. Steps: define specialist agents, add a triage agent with handoffs, run it with a session to keep context. Output: the final answer plus run history and a trace to inspect.
Repository summary in a sandbox
Input: a Git repository named in a manifest. Steps: create a SandboxAgent, attach a sandbox client, ask it to inspect the README. Output: a written summary of the project in the final output.

Before you choose

  • The SDK is free open-source software, but OpenAI model tokens, hosted tools and containers are billed separately.
  • It runs inside your own application, so you own deployment, state storage and tool implementations; a managed Agents API is a separate option.
  • The sandbox quickstart's local client supports macOS and Linux; Windows needs Docker or a hosted client.

Listed under

Get started

Install the package and set OPENAI_API_KEY to use OpenAI models.

  1. Create a Python 3.10+ virtual environment and run pip install openai-agents
  2. Create an OpenAI API key and export it as OPENAI_API_KEY
  3. Run the hello-world agent with Runner.run_sync, then add tools or handoffs
pip install openai-agents

Plans and prices

Prices in USD

No fee for the SDK itself. Model tokens, hosted tools and containers are charged under the OpenAI API rate table, and costs vary by model chosen.

OpenAI API usage (separate from SDK)Usage-basedSee the OpenAI API pricing table for current rates per model and tool.
  • Per-token model charges
  • Hosted tools billed per call or session

Source: developers.openai.com/api/docs/pricing, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

Ratings and reviews

No ratings for OpenAI Agents SDK yet.

Be the first to rate it

What’s changed

Free account

See what’s changed at OpenAI Agents SDK

Launches, price moves and pivots, each with a date and a source. Free with an account.

Signals

Free account

See OpenAI Agents SDK’s signals

Launch traction, hiring and release activity, tracked over time. Free with an account.

Data and security

Free account

See how OpenAI Agents SDK handles your data

Training on your files, compliance, hosting and the models behind it. Free with an account.

Compare with similar tools

Free account

Compare OpenAI Agents SDK side by side

Prices, platforms and features against similar tools, with sources. Free with an account.

Where this comes from

How we research
Company website
openai.github.io: product, pricing and policy pages
Web archive
When the site first appeared
YC directory mirror
Batch and company name
Web search
Confirms the official site
Social accounts
Accounts the company links, and its posts on X
Human review
Checked by a person
knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.