OpenAI Agents SDK

OrchestrationMIT

The OpenAI Agents SDK is a code-first toolkit for building agents that use tools, hand work to other agents and keep context across steps. The documentation recommends this path when you want typed application code in TypeScript or Python. The SDK runs inside your own application, with a runner that manages the agent loop. The documentation covers handoffs, guardrails, human review before risky steps, sessions, tracing for debugging and sandbox agents that run work in containers. Tools can be functions, MCP servers or hosted tools, and each language has its own open source repository.

OpenAI documents two paths: the SDK, which runs in your application, and an Agents API that runs a managed harness in OpenAI's service. This entry covers the SDK. Both repositories list the same core concepts, including realtime agents for voice.

Pinout

  • MCP server: not checked yet
  • HTTP API: not checked yet
  • command line: not checked yet
  • Python SDK: supported, verified
  • TypeScript SDK: supported, verified
  • OpenAPI description: not checked yet
Where each pin comes from
InterfaceStatusSource we readRead on
MCP serverNot checked yet——
HTTP APINot checked yet——
Command lineNot checked yet——
Python SDKSupported, verifiedAgents SDK quickstart, install
TypeScript SDKSupported, verifiedQuickstart, TypeScript install (npm install @openai/agents zod)
OpenAPI descriptionNot checked yet——

Facts

Facts about OpenAI Agents SDK
License (Python repository)MIT. Repository README (Python), read .
License (TypeScript repository)MIT. Repository README (TypeScript), read .
LanguagesTypeScript and Python, each with its own GitHub repository. Agents SDK overview, read .
Python installpip install openai-agents (also uv add openai-agents). Agents SDK quickstart, install, read .
TypeScript installnpm install @openai/agents zod. Agents SDK quickstart, install, read .
Python requirementPython 3.10 or newer. Repository README (Python), read .
TypeScript runtimesNode.js 22 or later; Deno and Bun listed as supported; Cloudflare Workers experimental. Repository README (TypeScript), read .
Latest Python releasev0.24.0, dated 2026-10-15 (marked Latest on the Releases page). Repository releases page (Python), read .
Core conceptsAgents, sandbox agents, realtime agents, agents as tools / handoffs, tools, guardrails, human in the loop, sessions, tracing. Repository README (Python), read .
Tool typesFunctions, MCP and hosted tools. Repository README (Python), read .
Deployment modelRuns inside your application; a separate Agents API offers a managed harness in OpenAI's service. Agents SDK overview, read .
Last commit. GitHub API, latest commit on the default branch of openai/openai-agents-python, read .

Call example

result = Runner.run_sync(agent, "Write a haiku about recursion in programming.")

Strengths

  • Both the Python and TypeScript repositories are under the MIT license.
  • The README lists handoffs, guardrails, sessions and tracing as built-in concepts.
  • Tools can be functions, MCP servers or hosted tools.

Alternatives to OpenAI Agents SDK

All orchestration tools for agents
  1. Open source agent framework with SDKs for Python, TypeScript, Go, Java, Kotlin.

    Orchestration
  2. Open-source Python framework for multi-agent workflows; separate commercial platform.

    Orchestration
  3. Low-level framework for stateful, long-running agent workflows built as graphs.

    Orchestration
  4. Durable execution platform, open source, used to run long-lived agents.

    Orchestration

Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check

Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.