TL;DR
MIT, Apache 2.0, BSD-3, fair-code, SaaS — what you can and can't do per framework license. Each row links to the vendor's own published documentation; we cite their stated capabilities and pricing rather than editorialising. See methodology for sourcing rules.
| Name | Vendor | Kind | License | Best for |
|---|---|---|---|---|
| AgentsBooks | AgentsBooks | OS / runtime | SaaS | service-firm operators running fleets |
| LangGraph | LangChain | library | MIT | stateful workflows; LangChain shops |
| CrewAI | CrewAI Inc. | library | MIT | business workflow automation |
| AutoGen | Microsoft Research | library | MIT | debate / decision systems |
| OpenAI Assistants API | OpenAI | API | proprietary | OpenAI-only agentic apps |
| Claude Agent SDK | Anthropic | SDK | MIT | Claude-centric agent stacks |
| n8n | n8n GmbH | workflow / no-code | fair-code | ops automations + glue |
| Zapier Agents | Zapier | no-code | SaaS | non-engineering teams |
Highlighted row = AgentsBooks. Comparison cells reflect each vendor's own published claims, accessed 2026-05-07. Rows are not editorialised.
Sources cited
- Independent comparison surveys — agent frameworks 2026. https://pecollective.com/blog/ai-agent-frameworks-compared/
- AgentsBooks — agentsbooks.com. https://agentsbooks.com/