#6767th of 142 in Agent frameworks and SDKs
agency-swarm
Python multi-agent framework built on the OpenAI Agents SDK
- Stars
- 4.6k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Aug 2026
- Language
- Python
4,185 on 22 Apr 20264,636 on 11 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Agency Swarm is a Python framework for building multi-agent applications on top of the OpenAI Agents SDK and Responses API. Agents get roles, instructions and Pydantic-typed tools, and an Agency object defines which agents may start conversations with which using directional communication flows (`ceo > dev`). Thread persistence is handled through save/load callbacks, and agencies can be run via a web demo, a terminal UI or async code.
Who it is for: Python developers building multi-agent systems with role-based agents
Strengths
- Explicit directional communication_flows control which agents can message each other
- Tools can be written with @function_tool, BaseTool or generated from OpenAPI schemas
- Non-OpenAI models (Claude, Gemini, Grok, OpenRouter) reachable through LiteLLM
- Built-in web demo, terminal UI and async API for running agencies
Weaknesses
- Tied to the OpenAI Agents SDK and Responses API; other providers go through LiteLLM
- Requires Python 3.12 or newer
- v1.x is a breaking rewrite from v0.x; existing projects need migration
- Persistence is manual: you implement the load and save thread callbacks
What it needs
- no GPU
- Needs OpenAI API key, OpenAI Agents SDK, LiteLLM (for non-OpenAI models)
- Models: OpenAI GPT-5 family, GPT-4o, Anthropic Claude via LiteLLM, Google Gemini via LiteLLM, xAI Grok via LiteLLM
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Agent frameworks and SDKs
See all 142| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 96 out of 100 |
| 2 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 92 out of 100 |
| 3 | mastraTypeScript framework for building AI agents, workflows and MCP servers | 92 out of 100 |
| #4 | browser-usePython library that lets an LLM agent control a web browser | 90 out of 100 |
| #5 | copilotkitReact and Angular SDK for adding agent chat and generative UI to apps | 90 out of 100 |