OpenAI Swarm

OpenAI's deprecated agent framework, replaced by the Agents SDK

Agents & Automation Open Source Open Source
Researched · Published · Reviewed
RECATOOLS Score
4.9 / 10
Capability
5
Value for money
5
Ease of use
7
ASEAN readiness
5
API quality
4
Founded
2024
HQ
San Francisco, California, USA
Users
Launched
Developer

Overview

Swarm was OpenAI's minimal, MIT-licensed multi-agent framework showing agent handoffs in under 1,000 lines of code. OpenAI replaced it with the production-ready Agents SDK in March 2025; Swarm is unmaintained.

Advertisement

Pricing

Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.

Free
Free
Free tier with core features.

Use cases

Multi-agent education Agent handoff patterns Research

What you can produce with OpenAI Swarm

  • MIT-licensed, open-source core under 1,000 lines
  • Agent handoff and 'routines' orchestration pattern
  • Reference implementation for building custom multi-agent systems
  • Superseded by openai-agents-python / openai-agents-js (production SDK)
Advertisement

ASEAN Perspective

OpenAI Swarm in Southeast Asia

ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).

RECATOOLS Verdict

Swarm was always meant as a teaching tool: a stripped-down demonstration of handing a conversation off between specialized agents without building a full framework around it. The ideas stuck — routines, handoffs, minimal shared state — and the source is still short enough to read in an afternoon if you want to understand the pattern. What matters for anyone evaluating it in 2026 is that it's dead as a build target: OpenAI replaced it with the Agents SDK (openai-agents-python and a JavaScript counterpart) in March 2025, and Swarm's own README now redirects new users there. The SDK carries the same handoff model plus guardrails, tracing, sessions and hosted tools, and is what OpenAI uses in its own reference examples and customer case studies. Existing Swarm code keeps running, since it's just calling the standard OpenAI SDK underneath, but there's no roadmap and no new model support. Don't start a new project on it.

Independent AI-assisted assessment by RECATOOLS.

What people say

Swarm shipped as an explicitly educational release — the GitHub description still calls it an "educational framework exploring ergonomic, lightweight multi-agent orchestration," managed by an OpenAI solutions team rather than a core product group. The pitch was minimalism: agents as simple Python functions, handoffs as a first-class primitive, no heavyweight abstraction layer, MIT licensed, under 1,000 lines of core code. For a few months in late 2024 it was the reference a lot of people used to learn how multi-agent handoff patterns work in practice.

That window closed in March 2025, when OpenAI released the Agents SDK as its production successor. The SDK kept Swarm's handoff concept but added the things Swarm deliberately left out — guardrails, tracing, session management, hosted tool integrations — and by May 2026 had reached version 0.17.1 with more than 26,000 GitHub stars, actively maintained and used in OpenAI's own examples and customer references. Migration guides comparing the two (Respan's "OpenAI Agents SDK vs Swarm" among them) are unambiguous about the direction: move to the SDK, don't build new work on Swarm.

Swarm itself hasn't been pulled — it's still on PyPI, and because it's a thin wrapper that just calls the standard OpenAI SDK underneath, old code keeps running as long as the underlying API doesn't change shape. There's no hard cutoff date. But there's also no new feature support, no bug fixes beyond whatever the community forks, and articles asking "is OpenAI Swarm still worth using in 2026?" reach the same answer every time: it's fine as a teaching reference for the handoff pattern, and a bad choice for anything you intend to run in production or maintain past a prototype. The honest way to list this entry is as a historical artifact that shaped how OpenAI (and others) think about agent handoffs, not as a live framework competing with LangGraph, CrewAI or the Agents SDK itself.

Summary of public user & expert reviews, compiled by RECATOOLS.

About this listing

Researched on
Published on
Last reviewed

This entry was compiled from publicly available data including OpenAI Swarm's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with OpenAI Swarm unless explicitly stated.

Data accuracy

Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.

For the latest details, please refer to OpenAI Swarm directly →

Spotted something out of date? Suggest an update →

Advertisement