AI Agent Frameworks
Agent frameworks provide the scaffolding to build LLM applications that reason over multiple steps, call tools and APIs, keep state or memory, and sometimes coordinate multiple agents. They differ mainly by language, how much control vs abstraction they give you, and whether they model workflows as graphs, roles, or plain code.
| AutoGen Microsoft's framework for multi-agent conversations. | Open source (MIT) | Self-hosted | Open source (free) | Python, .NET | Yes | Yes | Yes |
| CrewAI Role-based multi-agent orchestration in Python. | Open source (MIT) | Self-hosted, Managed | Open source (free) | Python | Yes | Yes | Yes |
| Dify Low-code platform to build LLM apps and agents. | Source available (Dify Open Source License (Apache-2.0 with conditions)) | Self-hosted, Managed | Free tier + paid | Visual + API | Yes | Yes | Yes |
| Google Agent Development Kit Google's open-source kit for building and deploying agents. | Open source (Apache-2.0) | Self-hosted, Managed | Open source (free) | Python, Java | Yes | Yes | Yes |
| LangGraph Build stateful, controllable agents as graphs. | Open source (MIT) | Self-hosted, Managed | Open source (free) | Python, JS | Yes | Yes | Yes |
| Letta (MemGPT) Stateful agents with long-term memory (MemGPT). | Open source (Apache-2.0) | Self-hosted, Managed | Open source (free) | Python, TS | Yes | Yes | Yes |
| LlamaIndex Data framework for RAG and agents over your data. | Open source (MIT) | Self-hosted, Managed | Open source (free) | Python, TS | Yes | Yes | Yes |
| Mastra TypeScript framework for agents, workflows, and RAG. | Open source (Elastic License 2.0) | Self-hosted, Managed | Open source (free) | TypeScript | Yes | Yes | Yes |
| OpenAI Agents SDK OpenAI's lightweight framework for agents and handoffs. | Open source (MIT) | Self-hosted | Open source (free) | Python, JS | Yes | No | Yes |
| Pydantic AI Type-safe agent framework from the Pydantic team. | Open source (MIT) | Self-hosted | Open source (free) | Python | Yes | Yes | Yes |
| Semantic Kernel Microsoft SDK to add AI orchestration to apps. | Open source (MIT) | Self-hosted | Open source (free) | C#, Python, Java | Yes | Yes | Yes |
Compiled from public docs and vendor sites; verify current pricing with the vendor. No vendor pays for placement. Each tool links to a fuller profile.
How to choose
For explicit, controllable stateful workflows use LangGraph or Google ADK; for RAG-heavy apps LlamaIndex is the most batteries-included. CrewAI and AutoGen focus on multi-agent collaboration. Pydantic AI and Mastra suit teams that want typed, minimal abstractions in Python and TypeScript respectively. Match the framework to your language first, then to how much abstraction your team wants.
FAQ
Do I even need a framework to build an agent?
No. Many production agents are plain code calling a model API in a loop. Frameworks help with tool orchestration, state, streaming, and multi-agent patterns, but they add abstraction you should be willing to learn and debug.
What is a multi-agent system?
An architecture where several specialized agents (e.g. planner, researcher, writer) collaborate, often passing messages or sharing state, rather than a single agent doing everything.
Are these frameworks free?
Most listed here are open-source and free to use; you pay only for the underlying model API calls. Some vendors offer paid managed platforms on top.