Best Open-Source AI Agent Frameworks (2026)
Open-source frameworks for building production AI agents, from graph orchestration to multi-agent crews and type-safe Python.
LangGraph
Low-level orchestration framework from LangChain for building stateful, multi-step agents as graphs, with persistence and human-in-the-loop support.
Open source (MIT) Python, TypeScript Official site
CrewAI
Python framework for orchestrating role-playing, collaborative multi-agent "crews" with tasks and processes.
Open source + paid platform Python Official site
AutoGen
Microsoft's open-source framework for building multi-agent conversational applications.
Open source Python, .NET Official site
OpenAI Agents SDK
Lightweight OpenAI SDK for multi-agent workflows with handoffs, guardrails and tracing.
Open source (MIT) Python Official site
Pydantic AI
Agent framework from the Pydantic team focused on type-safe, structured outputs and dependency injection.
Open source (MIT) Python Official site
smolagents
Hugging Face's minimal library for agents that think and act in code.
Open source (Apache-2.0) Python Official site
Mastra
TypeScript agent framework with workflows, memory, RAG, evals and integrations.
Open source TypeScript Official site
Google Agent Development Kit
Google's open-source toolkit for building, evaluating and deploying agents, optimized for Gemini.
Open source (Apache-2.0) Python, Java Official site
LlamaIndex
Data framework for connecting LLMs to your data, with agent and workflow abstractions for RAG-heavy agents.
Open source + paid cloud Python, TypeScript Official site
DSPy
Stanford framework for programming—rather than prompting—language models, with optimizers for prompts and weights.
Open source (MIT) Python Official site
Agno
Python framework (formerly Phidata) for building multi-modal agents with memory, knowledge and tools.
Open source + paid platform Python Official site
Letta
Framework (from the MemGPT team) for stateful agents with long-term memory.
Open source + paid cloud Python Official site