> For the complete documentation index, see [llms.txt](https://soyjakai.gitbook.io/soyjakai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://soyjakai.gitbook.io/soyjakai-docs/agentic-ai-frameworks/leading-agentic-ai-frameworks.md).

# Leading Agentic AI Frameworks

#### **LangGraph**

A graph-based framework designed for long-running, stateful workflows. LangGraph excels at approval-driven enterprise processes with durable execution and human-in-the-loop checkpoints.

Best for: IT and HR workflows requiring reliability and observability.

***

#### **CrewAI**

Built around “crews” of specialized agents, CrewAI focuses on role-based collaboration and task delegation.

Best for: Structured multi-agent task coordination.

***

#### **OpenAI Agents SDK**

A production-ready SDK for OpenAI-powered agents with tracing, guardrails, and session management.

Best for: Teams building tightly integrated OpenAI workflows.

***

#### **AutoGen**

A framework for delegated reasoning across multiple agents. Often paired with additional governance tooling for production deployments.

Best for: Complex agent interactions.

***

#### **LLaMA (Model Ecosystem)**

Not a framework itself, but a popular open-weight model family used as the reasoning layer within agentic systems.