Can it use non-OpenAI models?

Yes — via LiteLLM and any-llm adapters, it supports 100+ models, though OpenAI models are the default path.

Does it manage conversation memory?

Yes — Sessions automatically manage conversation history across runs.

SDKOpen sourceNot yet hands-on tested

OpenAI Agents SDK

Founded 2025

Build agentic AI apps in a lightweight, easy-to-use package with very few abstractions.

What it is

The OpenAI Agents SDK is a lightweight, open-source framework for orchestrating single- and multi-agent workflows around Agents, Handoffs, Guardrails, and Sessions, with built-in tracing. It is the successor to OpenAI's experimental Swarm SDK.

At a glance

7/7 facts sourced

Pricing: The SDK is free and open source (MIT); you pay OpenAI (or the chosen provider's) API/token costs.source ↗
Hosting model: A self-deployed library (Python, with a TypeScript version too) that runs in your own process and infrastructure.source ↗
Memory & state: Sessions provide automatic conversation-history management across agent runs — a persistent working-context layer.source ↗
MCP support: Nativesource ↗
Integrations: Built-in tools (web search, file search, computer use via the Responses API), function tools, MCP servers, agent handoffs, and tracing/observability.source ↗
Model providers: OpenAI-first (uses the Responses API by default for OpenAI models) but provider-agnostic — supports 100+ other LLMs via LiteLLM and any-llm adapters.source ↗
Open source: Yessource ↗

3.8of 5

Editorial scorecard

Overall 3.8 / 5

Our opinion — the same 0–5 rubric is applied to every platform, including Alfe. See the rubric.

Developer experience4/5

Pricing transparency & value5/5

Scalability3/5

Memory & state3/5

Integrations4/5

MCP nativeness4/5

Strengths

Minimal, low-abstraction API with clear primitives (agents, handoffs, guardrails, sessions).
Built-in tracing and native MCP support.
Free and open source (MIT), and provider-agnostic via LiteLLM.

Trade-offs

Optimized around OpenAI's Responses API — non-OpenAI providers go through adapters.
Library-only, so you run your own infrastructure.
A newer framework (2025) with a fast-moving API surface.

Our take

Editorial opinion — not a hands-on test result

In our opinion, the OpenAI Agents SDK is a solid, unopinionated choice for teams building agentic workflows in Python who value a small API surface and good tracing. It is most frictionless if you are already on OpenAI models; other providers work through adapters. As with any library, memory beyond Sessions and production infrastructure are yours to own.

FAQ

Can it use non-OpenAI models?: Yes — via LiteLLM and any-llm adapters, it supports 100+ models, though OpenAI models are the default path.
Does it manage conversation memory?: Yes — Sessions automatically manage conversation history across runs.

Compare OpenAI Agents SDK

Sources

OpenAI Agents SDK docs · accessed 2026-07-01
OpenAI Agents SDK on GitHub (MIT) · accessed 2026-07-01
OpenAI launch announcement · accessed 2026-07-01

The Agent Examiner editorial · last updated 2026-07-01. Facts are drawn from each vendor’s own public materials; scores are our editorial opinion (methodology).

Product and company names are trademarks of their respective owners and are used here only to identify the products under discussion.